Home

The LyX User's Guide - University of Adelaide

1. Figure 4 2 Geometric definitions of a disc brake system Shigley and Mischke 2004 p 830 The material of the brake pads is a dry sintered metal The coefficient of friction and maximum allowable pressure of this material are u 0 3 and pa 2MPa respectively Shigley and Mischke 2004 p 842 The maximum allowable pressure of this material relates to its ability to withstand pressure without being damaged The available torque from the chosen brake system is 1 T 5 x 18x im x 0 3 x 2 x 10 x 0 054 x 0 07 0 0542 4 2 T 10Nm 4 3 This torque value represents an upper limit that the brakes are capable of applying to each drive wheel The brake calipers chosen operate using a cam mechanism for which data is not available This makes analysis of the actual braking capability untenable The upper limit of the braking torque is then an estimate only and is on the same order as the maximum torque applied to the drive wheel by the motors 22 4 2 Brake Lever Brake cable modification Figure 4 3 Hand operated cable lever chosen for use in EDWARD 4 2 Brake Lever The mechanical brake is to be operated by the occupant A foot operated brake which was con sidered in the diwheel preliminary report Dyer et al 2009 was determined to be unergonomic due to the restricted movement of the occupant s legs A single hand operated brake lever was then chosen to operate the cable drawn brakes The brake lever chosen was a mot
2. High stresses in the outer wheel when compared to half the yield stress of steel indicated the possibility of local deformation in the event of an impulse overload e It was recommended that thicker walled piping be used for the outer wheels this will also increase the inertia of the outer wheels making the inner frame more susceptible to slosh 6 2 3 Space Frame Analysis FEA was performed on the structure of the space frame The purpose of this investigation was to determine how the space frame would distribute loading when in operation to identify areas of concentrated stress and to assess the viability of the design The material of the space frame modeled was 25mm OD steel tubing with a 2mm thickness The properties of steel used were a Young s modulus of E 200GPa Poisson s ratio of v 0 3 and a yield stress of a 250MPa The loading of the space frame considered was the static weight of one passenger and other components overestimated to be 200kg These loads were applied at two points at the location of the seat mounting A half model of the space frame was created in ANSYS using the plane of symmetry between the wheels of the diwheel The model was also simplified by not modeling the details of each joint in the space frame A contour plot of the VON MISES stresses of the solved FEA model is shown in Figure 6 12 6 2 4 Space Frame FEA Results and Conclusions The results obtained from the solved FEA model of the space frame ar
3. Is encoder Right second pulse zero PTT1 0 Is encoder Left second pulse zero PTT6 0 N Yes E Left is anti clockwise Right is anti clockwise Left is clockwise Right is clockwise Right Hz number of pulses Left Hz number of pulses divided by time divided by time Figure 8 2 Flowchart of the algorithm used to measure encoder speed 111 8 Software Design Figure 8 3 Sensor board combining a gyrosensor and two axis accelerometer The accelerometer measures the projection of the gravity vector onto its two axes as shown in Figure 8 4 As the inner frame rotates with respect to the gravity vector a larger reading will be measured on either the x axis or y axis of the accelerometer The angle estimate of the inner frame is then given by accelerometer gree atan 2 8 1 accelerometer x While the arctangent function can correctly estimate the inner frame angle based on the pro jection of gravity onto each axis it suffers increased noise when the magnitude of the readings on either axis becomes small i e near 0 and 90 For this reason the accelerometer should be orientated at 45 at the neutral position lower equilibrium if possible Unfortunately the accelerometer estimate for 0 is affected by noise since the accelerometer cannot distinguish be tween the static acceleration of the gravity and dynamic acceleration of the
4. e Estimating the masses of the wheels to be m 40kg and frame and passenger mp 200kg giving a total mass of m 240kg e KE is the total kinetic energy of the diwheel e K Erotat is the rotational kinetic energy of the outer wheels e K Evans is the total translation kinetic energy of the diwheel e KF Ff is the translational kinetic energy of the inner frame e K Ew is the translational kinetic energy of the outer wheels v 8km hr 8 3 6m s 2 22m s is the translational speed of the diwheel e R 0 75 is the radius of the outer wheels e Iw is the moment of inertia of the outer wheels Iw my R 40 0 75 22 5kgm e PE is the gravitational potential energy of the inner frame e h is the distance of the center of gravity from the center of rotation of the diwheel 158 B 3 Mechanical Design Dynamic Analysis The total kinetic energy is given by KE K Evotat Arana re oer KEp K Ey trans Which in a braking situation is converted to potential energy PE m g9 he gt K Erotat KEp f K Ew eka PE 1 1 1 gt 7 wey meo swe trans Me 9 he 1 1 1 1 gt heg a Sivek mpu smwe T 1 1 1 1 E 5 x 2 22 0 75 52 2 22 4 2 222 o gt hog says 5 225 0 75 5200 x 540 x X gt heg 0 294m With the center of gravity at this location a locked braking situation at this speed will fully invert the diwheel into an unstable equilibrium which is then subject to
5. kikeNns 7 36 p s 0 0247s 0 00105s 0 3101 737 V s 53 0 9903s 7 124s 1 977 7 87 which has a steady state gain of 0 3101 1977 0 1569 rad s volt This is equivalent to a top speed of 24 x 0 1569 x 0 72 x 3 6 9 76 km hr at 24 Volts This is quite different from the top speed predicted simply due to the back EMF of the motor which predicts a top speed of 3100rpm x 27 x 0 72 x 3 6 5 14 x 7 x 60 23 39 km hr This large discrepancy is caused by the large damping due to the differential motion of the inner frame inside the wheels principally due to idler friction and is likely to decrease as the system is bedded in Speed d 2 K voltage 0 02475 0 00105s 0 3101 p s 0 99035 7 124s 1 977 Figure 7 9 Block diagram of open loop speed control 7 4 Yaw Control Yaw control is required to manoeuvre the diwheel Open loop and closed loop yaw control provide feasible options for controlling the yaw rate Each option uses a differential drive system to create a yaw rate However a closed loop system uses feedback to minimise the error between the desired yaw rate and actual yaw rate The yaw rate is determined by the difference in the wheels angular velocities divided by the distance between them 91 7 Control System Design lt OR Figure 7 10 Schematic of the diwheel used in yaw dynamics derivation ROL PR 7 38 L by providing equal but opposite incremental voltages to each mot
6. 0 is related to the change in height of the CoG of the inner frame and is given by E eg my 1 cos 09 7 13 where g is the gravitational acceleration Lagrangian Now with the total kinetic and potential energy of the system the Lagrangian Ej Ep may be found a 2 2 2 5 Zm Ere g P E Rema cos 0 p0 4 z 6 e g ma e g ma cos 0 7 14 76 7 1 Analytical Derivation of the System Dynamics It is desirable to reduce the size of this equation to make further maths operations neater This may be done by collecting the constants before state variables and replacing each of the expressions with a single coefficient Performing this simplification yields A A L ay ar cos 0 0 ze ay cos 8 1 E where KA J R mi ma is the effective moment of inertia of the wheel and inner frame about the contact point with the ground Ja Jy e ma is the effective moment of inertia of the inner frame about the centre of the wheels with regard to the coupled motion of the inner frame and wheels and ag R e ma and ag e g ma are constants of convenience Euler Lagrange equations With the Lagrangian determined the Euler Lagrange equations may be applied to yield the system dynamics in the form d OL OL ae dt 5 90 bia 0 TE 7 16 d OL OL eo Tt 7 dp bialp 0 b16 T 7 17 where by is a viscous damping coefficient relat
7. e The energy associated with the rotational velocity of the drive wheel is considered negli gible e There is no slip between the drive wheels and the outer wheels e There is no slip between the outer wheels and the ground 7 1 6 Model Parameters The parameters used for the model and thus the controller designs are detailed in Table 7 1 Most parameters were estimated from the Solid Edge model of the diwheel and driver with the exception of the damping terms which were measured The motor parameters were source from supplier data 7 2 Simmechanics Derivation of the System Dynamics In order to verify that the dynamics of the two degree of freedom DoF found in Section 13 are correct a model of system was also completed using the SIMMECHANICS toolbox in MATLAB In comparison to the energy based analytical method of Section 13 the SIMMECHANICS model derivation utilises a block diagram approach to describing the system dynamic behaviour which requires e The identification of separate bodies with mass and inertia e Defining relationships between each body DoF using joints e The measurement of joint body displacements velocities and accelerations 81 7 Control System Design Table 7 1 Parameters used to define the model Note that the terms for the wheels and motors account for both acting together Part Parameter Value Wheels mi 50 3 kg Ja 26 1 kg m Frame ma 218 kg Ja 48 4 kg m Lengths R 720 m
8. is the force produced at the ground by the differential velocity of the wheels e Kk is a gain which determines how much of the system voltage is reserved for yaw motion The yaw angular rate is given by Xx Ya bas Differentiating gives ia Rx PA q d The sum of the moments for the wheel is given by Jwa Ba Rx Fy T The sum of the moments for the diwheel is given by Jy dx Fy 0 Combining Equation 7 41 and 7 42produces Jupa OA ee ed Jy R R R Substituting Equation 7 40gives 1 0 2 5 9 lt The differential torque of the motor is given by combing the torque of both motors kVa keki NN i 2N N T Ro Combining with 7 43 produces E 8 s 2 4 a 2 rason 8 aa Taking the Laplace transform produces v s k N N 1 Va s Rm 8 Jy 4 Jw 82 B 2keki N2 N23 s Differentiate to give the yaw rate 7 39 7 40 7 41 7 42 7 43 7 44 7 45 7 46 93 7 Control System Design Step Response yaw output rate rad s 0 5 10 15 20 25 time sec Figure 7 11 Open loop yaw response for a 14 4V step input v s ki N N 1 Va Rn 4 Jy 4 Jo s B 2 2kekiN2N2 4 wat Substituting values for the estimate system parameters produces pls _ 12 Va s 92 158 22 42 7 48 7 4 2 Open loop Yaw Step Response The open loop dynamics given by Equation 7 48 have
9. m mis m s 2 Initial Condition Translational Sensor th dth d2th rad radis rad s 2 Rotational Sensor Figure 7 3 Inner frame SIMMECAHNICS block diagram gt 2 To Outer Wheel To Inner Ring l f 6 E Drive wpheel Initial Condition Inner Ring Drivewheel gearing Joint Joint Actuator Disturbance Torque 2 E 4 Drive wheel torque Inner Ring senso Outer wheel torque Nm Outer wheel torque Sprocket j gearing Bd Motor Output Shaft Torque Drivewheel Viscous Friction El Motor Gearbox Input voltage 4 Motor Torque Inner ring Speed 1 O L A vi A Motor armature torque AJ ting soale Drivewheel e Resistance Current Limit Torque Angular velocity Constant back emf Sprocket Motor Back EMF gearing Gearbox1 Constanti Motor Current Amps Drivewheel Angular Velocity rad s Integrator A Drivewheel Angle rad Figure 7 4 SIMULINK block diagram of the drive system including motor dynamics 84 7 2 Simmechanics Derivation of the System Dynamics A __ _ lt Body Actuator Sal Pa Wheel Linear alcs4 q cs3fs S Selector Velocity mis CO csi cs2B ToMotir Body Sensori u Y Big Wheel Wheel Angular Velocity rad s Selector4 Rolling Constraint Ground1 Figure 7 5 Block diagram of the combined outer wheels sigh 7 2 4 Outer Wheels Because both the ana
10. where the Lagrangian L is an expression of the difference in the kinetic and potential energies of the system qi are generalised coordinates in this case 0 and w and F are generalised forces The solution of the Lagrange equation for each coordinate qi yields an expression of the form M a C q 4 G q F 2 which summarise the system dynamics Velocities The translational and rotational velocities of the bodies comprising the diwheel are presented below in prepara tion for the Lagrangian The translational velocity of the wheel body 1 in the z direction is Uir RE 3 where R is the outer wheel radius and is the angular velocity of the wheel The translational velocity in zx direction of the CoG of body 2 the inner frame is Voz R e cos 0 4 where e is the eccentricity between the inner frame CoG and the centre of the wheels and is the angular velocity of the inner frame relative to an earth centred frame The corresponding velocity in y direction of the inner frame CoG is vay Oe sin 0 5 The magnitude of the velocity of the inner frame CoG is thus 2 loa Re de cos 0 6 e sin 0 6 Kinetic Energy The kinetic energy of the diwheel has been separated into the following terms First the rotational energy of the wheel SP TA 7 where J is the combined moment of inertia of both wheels about their centre Second the translational energy of the wh
11. 00000000000 eee Brake levers seca ha a lote a ell a A ee Se eh we Pela e Outer Wheel a ciie eae sa ee he a doo as ho Seating and Harness 624 2 Sood A BUS ae Bis a axel eds Drive Wheel ii A a eer nla a ate g thee sled ee o dente 4 1 4 2 4 3 4 4 4 5 4 6 Motors vii Contents AR Motor Controller e Li A e A ii A o A dy eo Ae hed AS Denied ee A fe Bot See A eee Os Eh re en k 4 9 Processor 22 8 bs O AO A ee ele 410 Gyrosensor Vig et archers i ne hee ee ee a oe eel a whe as ke a 4 11 Accelerometer de aN a Oe Bl ge a Be se a hee tinal ag 4 12 Steering Control es ets as cat aif aes Ott ok de E a Bo BOS BR Peep hg 5 Electrical Hardware gel Motor Controller 24 k srs abe ft OR be PE oh SA OGY A DE PEED SHEA 32 dower Board a s Ave Bice eh BOF ek BB ie Bo oo BNO eee tee E Bee 5 3 Power Control Board was ust Punk uyr Pr Mth ten eth ha eth Le tb Any GA OTA ON A Gow OX eva acne ae ate pak by wera bn Sk ee wicks Goo DOUETeITE e A Sek oe ok BB ee E Se Bela Boe ee es 5 6 Gyrosensor and Accelerometer Board Sensor Board ST Encod r Sens rs r ot lt 6 Tea Wik eS ee Be we SE ee Se hc ae Baek 6 Mechanical Design 6 1 Space Frame Design tata EBA Be RS ots NS ot de dd o OL Dynamics ed za br igh Send Nigh Poet ARs eA aia dots Ae ete dd 04 25 HRPOnOmics fiat a one at neva ih whee des ela des o ice D ee cee ae 6 1 3 Final Design of the Space Frame soi ad eos 61 4 Electr nics OK acto ods een ky ene fel eR Ba
12. 116 8 6 2 as A O ea a Da ea eG BS 2G AG 117 Sr Additional P nctions e esea A SS AAA a da da 117 SL sine the SL Gard lala 2 aa tu a A A AS A BA 117 9 Experimental Results and System Integration 119 9 1 System parameter identification o oo a e ee 119 OZ VON SUE Tests Saer sheeted wy a iar gee hn ae ot eee enia E a as rna nE a 120 9 2 1 Open loop slosh bh a Se Be a 121 9 2 2 Closed loop slosh control le pio a e a a 121 9 2 3 Positive Feedback Swing up Controller Test 124 9 3 Software and Electrical Issues 0 la a ia ed RRA 125 94 Hardware ssues a a a a 126 10 Final Design Analysis 127 10 1 Summary of Goals and Achievements a e 127 10 1 Mechanical Goals a 2 2 ie a lt ae we ee ete er 127 Xi Contents 10 1 2 Electrical and Electronic Goals 10 1 3 Miscellaneous Goals a a 10 1 4 Extension Goals a a MN O A A Dede peed Os Sa el ag aa 11 Summary TEM Future Work wi irsi seg rs eee a eR ee Be ce a HR da TP CONCISO s ef nd tay oof A Oh Peet Oh te A dts dee a ares ee References A Publications Arising from this Thesis B Appendix B 1 Hertzian Idler Wheel Contact a a each ee eae Bk B 2 Dive wheel Hub Bolsas eee Das a a RW a eth de ats eae B 3 Mechanical Design Dynamic Analysis 00000004 Bal Anthropometris dosa a Keak e ie teas At Mid Gs Ad AA GB he C FMEA Risk Analysis and SOP D Mechanical Drawings
13. 24 4 4 Seating and Harness Outer Wheel Tyre The tyre of the diwheel is required to provide traction and vibration isolation between the wheel and the road It also eliminates the rigid rolling contact between the wheel and the ground The tubular wheel chosen requires a compliable surface to be attached to the outer radius The compliable surface must provide grip and wear resistance and must bond well to a metallic surface From email correspondence with ALLIED RUBBER on the 23 March 09 natural rubber was recommended as the most feasible option As budgetary constraints limit the selection of many components sponsorship was sought while looking for potential rubber suppliers Sponsorship was recieved from RUBBERTEX PTY LTD for the adhesion of 6mm thick rubber strips to the outside of the wheels The method used by RUBBERTEX was to firstly prepare the outer surface of the wheels by buffing them A steel primer was used before two coats of an adhesive The rubber strips were pre cut from a sheet of a product called Chemoline 4 Chemoline 4 is a Butyl based synthetic rubber which has a bonding skin layer which is revealed when a protective film is peeled away The rubber was applied as one continuous length and joined at its ends using a lapped joint Figure 4 6 Figure 4 6 The lapped join of the continuous rubber strip of the outer wheel tyre 4 4 Seating and Harness The occupant is required to be constrained in various angles of rotat
14. 5 05 J y 3 3 0 7 jo 0 5 L L l l l 1 i i f f 0 1 2 3 4 7 8 9 10 Time seconds Figure 9 3 Closed loop slosh control with proportional feedback K 31 Closed loop step response for 6 Kp 62 0 5 T T T T T T T T T T 0 El T 0 5 4 14 4 1 5 1 1 0 1 2 3 4 5 6 7 8 9 10 Time seconds Closed loop step response for 6 Kp 62 0 5 T T T T T T T T T T T 3 g gt 0 y D lo 0 5 i 0 1 2 3 8 9 10 5 6 Time seconds Figure 9 4 Closed loop slosh control with proportional feedback K 62 123 9 Experimental Results and System Integration Closed loop step response for Kp 10 0 5 T T T T T T T T T T 6 rad 0 1 2 3 4 5 6 7 8 9 10 Time seconds Closed loop step response for Kp 10 SS Aaaa TAA r TA Time seconds Figure 9 5 Closed loop slosh control with proportional feedback K 10 9 2 3 Positive Feedback Swing up Controller Test An earlier hypothesis suggesting that changing the sign of the slosh controller gains would cause the system to go unstable and hence swing the inner frame up was inadvertently tested during the testing of the slosh control The required sign of the negative feedback gain was initially difficult to discern due to the opposite directions of the input torque and the movement of the inner frame and the dependence on the orientation of the gyrosensor For this reason the r
15. 7 Control System Design Table 7 2 Fuzzy rules Rule 1 Hf 0 PS 0 PS then Vm NB Rule 2 If 9 NS amp 0 NS then Vm PB Rule 3 If PS amp 6 NB then Vm PB Rule 4 If 0 NS amp 0 PB then Vm NB Rule 5 If 0 PM amp 6 PB then Vm NB Rule 6 If NM amp 0 NB then Vm PB Rule 7 If 0 PM amp 0 PS then Vm NB Rule 8 If 0 NM amp 0 NS then Vm PB Rule 9 If PM amp 0 NB then Vm NB Rule 10 If 0 NM amp PB then Vm PB Rule 11 If 9 PB then Vm PS Rule 12 If 0 NB then Vm NS Rule 13 If 0 PB 0 PB then Vm Z Rule 14 If 9 NB 9 NB then Vm Z 7 6 2 Balancing Controller The balancing controller is used to capture the inner frame as it is brought to within 10 degrees of the inverted position The balancing controller has been designed as a basic full state feedback regulator This allows all of the poles of the open loop system at the inverted position to be moved into the left hand side of the complex plane The effect of relocating any of the unstable poles is to provide a stable response to any impulses exerted on the inner frame when it is inverted This section will describe the control system used to balance the inner frame upside down as well as application to the diwheel itself Full state feedback regula
16. 95 7 Control System Design Step Response yaw output rate rad s 0 2 4 6 8 10 time sec Figure 7 13 Closed loop yaw response for a 14 4V step input 7 5 Slosh Control Slosh control forms one of the principle project goals and aims to reduce the large amount of rocking sloshing of the inner frame which is an intrinsic property of the diwheel s open loop dynamics By limiting the sloshing motion of the inner frame the driver will be able to experience a much smoother ride As mentioned in Section 2 3 previous methods of actively controlling slosh have generally employed some form of a Proportional Integral Derivative PID control using feedback of some of the states of the system Examples include the 2DoF slosh rig Gandhi et al 2009 and the ball and hoop system Wellstead and Readman Proceeding in a similar fashion to these implementations of slosh control a form of PID controller has been designed for the diwheel During operation of the diwheel the centre of gravity must be displaced by some angle slosh angle 0 in order for forward or reverse motion to occur During harsh acceleration and braking the inner frame will slosh and oscillate until it reaches a steady state angle This steady state angle will be non zero in the case that the diwheel is to be driven at constant non zero velocity as a reaction torque supplied by the offset centre of gravity is required for the diwheel to overcome the external
17. However the driver of the diwheel attempted to maintain a constant input command by holding the joystick fixed From experience it was estimated that approximately 30V was applied to the motors 9 2 2 Closed loop slosh control The second test included a basic proportional controller using negative feedback of the slosh angular rate as described in Section 7 5 The slosh angular rate was measured by the gyrosensor in the same manner as the open loop testing It should be noted that as the sign of the input torque and slosh angle are opposite the feedback gain was implemented as positive rather than negative However the effect of the gain was that of negative feedback To discern whether the slosh controller gain tuned from the simulated model of the diwheel was applicable to the physical system three different controller gains were tested The results of the closed loop slosh controller for these three feedback gains are given in Figures 9 3 9 4 and 9 5 It can be seen from comparison of the closed loop responses for each controller gain that the simulation tuned controller gain of K 31 gives the best results This gives further evidence of the ability of the simulated mathematical model to accurately describe the physical system s behaviour The maximum slosh angle from Figure 9 2 was 1 2 rad Comparing this result to Figure 7 16 allows a comparison of the dynamic model with the physical system to be made Because Figure 7 16 was a step re
18. Virtual simulation model A virtual simulation shall be created for the MATLAB environment and operating features of the vehicle shall be reproduced in this simulation A virtual model of the diwheel was created to provide visual feedback of the modelled sys tem s response to command inputs The final VRML model utilises the dynamics derived for the mathematical model and gives a good estimate of the behaviour that the physical system may exhibit from commands in open loop slosh and inversion control modes The VRML model takes commands from a joystick identical to that used on the diwheel and the design strived to match the simulated model to the physical system Section 7 2 7 depicts the final VRML model and environment which allows operating modes to be controlled and switched between in the same manner as on the diwheel The completion of a fully functional VRML model demonstrates that this goal has been achieved 10 1 4 Extension Goals Regenerative Braking The kinetic energy of the di wheel at speed shall be returned to the batteries during braking if feasible The dynamic braking introduced automatically by the motor controller is believed to auto matically provide charge to the batteries causing regeneration This operation is guaranteed if 130 10 1 Summary of Goals and Achievements the motor controllers short circuit the motor when driven however regeneration has not verified during system testing State Control of Pitch An
19. concludes the report and lists possible future work which appears relevant to continuing the success of EDWARD 2 Background A literature review is essential for the design of an electronic diwheel as it creates a basis for system integration part design and component selection The following chapter presents a review of a number of which demonstrate what has already been achieved towards the design of a diwheel 2 1 Previous Designs From a review of existing diwheel s mechanical components a report and assessment of systems in previous diwheel designs was conducted Previous monowheels which use similar systems to that of a diwheel were also considered An internet based academic search revealed no literature directly relating to the mechanical design of a diwheel however there exist many designs outside the formalisation of academia For the purpose of the discussion of the mechanical system a diagram labelling the relevant components in shown in Figure 2 1 The term inner frame will be used to refer to the inner wheel and space frame which together constitute all components which rotate inside the outer wheels Figure 2 1 Components of a diwheel Modern designs of diwheels and monowheels have been used only on a small non commercial scale as backyard projects and attractions Perhaps the closest use of the diwheel principle in commercial applications is the prototype of Andr Costa winner of the 3rd Peugeot Design Competiti
20. i e when and where hazard is present task activity Hair Jewellery or Clothing may be entangled by the sprockets and chains Hierarchy of Control O El O Su C En Is IX Ad C PPE Current Controls Requirements for operators Clothing must be tight fitting b Long hair must be confined close to the head by an appropriate restraint b Finger rings and exposed loose jewellery eg bracelets and necklaces must not be worn Medic Alert bracelet must be taped if exposed Action Required YIN N Tfr to CAR YIN Can the following items become entangled eg in moving parts NA Are Emergency Stop buttons adequate q Yes within easy reach and NA NA NA Two Emergency Stop buttons are C El Emergency Stop buttons are N NA clearly marked installed O Su tested before all operations O No One manually activated Stop button amp En O oth under joystick controller Si One automatic micro switch on O is SSUES mechanical brake C Ad C NA C PPE Page 3 of 10 Project No 757 Issue No 1 Ss THE UNIVERSITY a F ADELAIDE IEE Ei PROJECT RISK ASSESSMENT Hazard Identified Likelihood Consequence Score Comments i e when and where hazard is present task activity Hierarchy of Control Current Controls Action Required YIN Tfr to CAR YIN Can anyone be crushed by X Plant falling or unexpected Pos
21. over the motors As the ROBOTEQ controller contains an internal processor several system parameters may be set by flashing the controller s internal memory using a dedicated computer program Such parameters as maximum current and acceleration rate could be customised in the program However because the project only required a power amplifier most of these parameters were zeroed or set to values that removed intelligent operation from the controller WRobotea DC Motor Controller Figure 5 3 ROBOTEQ dual H bridge motor controller Picture from Clark et al 2005 39 5 Electrical Hardware 5 2 Power Board Power to the vehicle is provided by the series connection of three 12V batteries The initial design featured the series parallel arrangement of the batteries to provide a larger current supply at 24V However simulations involving the inversion control scheme described in Section 7 6 showed that the lower system voltage of 24V made the swing up of the inner frame harder to achieve Some of the advantageous features of a series parallel arrangement of the batteries such as even power consumption easy recharging and simple voltage monitoring continue to apply however the higher bus voltage was taken as a preference in the selection of the battery arrangement Power to the system is provided by a Power Board which provides the numerous required system voltages The diwheel system requires the following working voltages e 9V reg
22. puts appear appropriate for the problem of swinging the diwheel up Membership functions The membership functions of the two inputs are shown in Figure 4 and are simi lar to those presented in Martynenko and Formal skii 2005 The range of is 7 7 and the range of is 27 27 The output voltage membership func tion consists of five singleton sets corresponding to 48 10 0 10 48 Volts d0 dt Figure 4 Membership functions Rules The rules chosen for our initial fuzzy controller have been designed to force energy into the system in or der to bring the inner ring to within 10 of the inverted equilibrium point and are detailed in Table 2 The ba sic premise is to swing the inner ring hard until the back EMP of the motor develops sufficiently and then to drive the motor so that it reinforces the fall of the inner ring due to gravity For small angles 0 lt 7 2 the action of the motor is determined according to the direction of the angular velocity with regard to the fact that a pos itive motor voltage produces a negative reaction torque on the inner ring For medium angles 1 2 lt 0 lt 37 4 the motor is also driven hard when the angular velocity causes the inner ring to approach the balancing point As the direction changes the motor direction is reversed When the angle becomes large 37 4 lt 0 lt m and the angular rate is small the motor action is made to cause the diwheel to drive u
23. to the current sensor 43 5 Electrical Hardware VBATTERY 10 Boy DIFFERENTIAL 10m OUTPUT 2 5V FS 10A FS Figure 5 7 Differential output using LTC6103 Left Single ended output using LTC6104 Right Munson 2006 The output from the current sensor board is a voltage between 0 5 Volts representing the current flowing Performance of the current sensor is shown in Figure 5 9 Figure 5 8 One of the current sensor boards 5 6 Gyrosensor and Accelerometer Board Sensor Board The gyrosensor and accelerometer have been combined together on the sensor board to form a simple attitude reference system as shown in Figure 8 3 This board simply contains the two breakout boards of the gyrosensor and accelerometer described in Section 4 10 and Section 4 11 as well as a number of headers which connect directly to pins on the two breakout boards The red DIP switch is used to select the sensivity of the accelerometer Use of the sensor board on EDWARD is described in Section 8 5 1 44 5 6 Gyrosensor and Accelerometer Board Sensor Board Response of the current sensor 2 51 Output voltage Volt o 40 20 0 20 40 60 Motor current Amp Figure 5 9 Performance of the current sensor board The mounting of the sensor board needed to take into consideration the axes of operation of each sensor The accelerometer measures acceleration on three orthogonal axes and as such could be mounted in any orien
24. 2 Swing up Control By simply altering the sign of Equation 8 13 the system becomes unstable and automatically swings the inner frame upside down 8 7 Additional Functions A number of additional functions have been implemented on the Dragonboard They include e A basic button driven menu system e Functions to use the SD card circuits provided on the board 8 7 1 Using the SD card Because of the delays faced during the project the implementation of a flexible data logging method on the microprocessor has not been fully completed However functions for reading and writing to the SD card were implemented and successfully tested The SD card is used to provide the system with an almost unlimited amount of non volatile storage though typically 1GB is all that is envisaged at this stage because higher capacity cards have not been tested The SD card protocol comes in two varieties however only one is ever used without proper licensing Such a protocol is standard SPI The Dragonboard provides SD card interface circuits which connect to PM2 PM3 and PM4 providing Write Enable Card Detect and Chip Select input outputs The card is written to and read from using the SPI on Port S using pins 4 5 and 6 The process of reading and writing to the card is as follows e Ensure a card is inserted check that PM3 is low e Initialise the card by running SD_ init and ensuring it returns 0 e For writing populate the global array buffer wit
25. 2005 The measurement of instantaneous angular speed Mechanical Systems and Signal Processing Vol 19 pp 786 805 Lygouras J 2000 Accurate velocity evaluation using adaptive sampling interval Micropro cessors and Microsystems 24 265 269 Lyons R G 2005 Understanding Digital Signal Processing Pearson education Asia Limited and China Machine Press Martynenko Y 2007 Motion control of mobile wheeled robots Journal of Mathematical Sciences 147 2 6569 6606 doi 10 1007 s10958 007 0496 4 140 References Martynenko Y and Formal skii A 2005a Theory of the control of a monocycle Applied Mathematics and Mechanics 69 4 516 528 doi doi 10 1016 j jappmathmech 2005 07 003 Martynenko Y G and Formal skii A M 2005b Theory of the control of a monocycle Applied Mathematics and Mechanics 69 516 528 Modaressi Tehrani K Rakheja S and Stiharu I 2007 Three dimensional analysis of transient slosh within a partly filled tank equipped with baffles Vehicle System Dynamics 45 525 548 Monowheels 2009 The museum of retro technology Viewed 12 March 2009 Munson J 2006 Dual current sense amplifiers simplify h bridge load monitoring Technical report Linear Technology Muskinja N and Tovornik B 2006 Swinging up and stabilization of a real inverted pen dulum IEEE Transactions on Industrial Electronics 53 2 631 639 doi 10 1109 TIE 2006 870667 Nichkawde C Harish P M a
26. 4 4 A Hall Effect sensor used in sensing the operation of the brake lever scale approx imately 15mm lengthwise 4 3 Outer Wheel The outer wheels of the diwheel provide rolling support of all other components of the diwheel They also determine the maximum outer diameter of the diwheel The outer wheels mate with the idler wheels and are in rolling contact with them Some common solutions for a diwheel s outer wheel are reviewed in Section 2 1 A pneumatic outer wheel was determined to be unfeasible as a suitable component does not exist and would be expensive to manufacture Subsequently the solution of rolling a tubular wheel was chosen Analysis of the strength of these wheels in rolling contact with the idler wheels can be seen in Section 6 2 As the outer wheels must be in rolling contact with the idler wheels the outer wheel must be partially exposed If this section were coated with paint in order to prevent corrosion it is likely to be worn by the idler wheels and would crack and fail Because of this the outer wheels were chosen to be rolled from stainless steel piping 50mm nominal bore 60 3mm OD schedule 40 wall thickness of 3 91mm The piping was rolled to an outside radius of 1500mm and cut and welded to form a ring The completed outer wheels with a rubber tyre adhered to the outer surface are shown in Figure 4 5 Figure 4 5 The finalised assembly of rolled stainless steel tube and rubber tyre to form the outer wheel
27. Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 1 LENGTH 04 TITLE Inner rotational frame member short UNLESS OTHERWISE SPECIFIED SCALE 1 1 QUANTITY 6 Quantity 8 S Please make the 37mm hole for interferencely fit for 37mm OD and 12mm ID deep grove ball bearings bearing supplied r p Chamfer with 1 mm setback for installing bearin Sen Cazzolato DIWHEEL WHEELBIKE A i pe Lanio SUPERVISOR Ben Cazzolato CONTACT 0413647219 DETAIL A TITLE Single bearing housing UNLESS OTHERWISE SPECIFIED QUANTITY 8 make this item with 35mm side 3mm thickness SHS Quantity 8 DRAWN 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Suspension bridge member sides Steel TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 6 QUANTITY 8 make this item with 35mm side 3mm thickness SHS Qua ntity 4 DRAWN 05 27 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Suspension bridge member top Steel TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 6 QUANTITY 4 weld bot DRAWN 05 26 09 CHECKED Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED Quantity 2 DIWH
28. Dynamics solureuAp Ste Iouur pue MeA pardnosun y SUIPNOU popow yunus eur 8 Oms IPPON TAA dp isd y1yd 714d eyyy AS zs pow jouog apow oywos ndu a eyga suoyng apow jouog BIGesigsa qeug VOISIBAU agesiq aiqeuaysols uopngaeig uoynganug 1 o i SWE U MEA sapow jouog Iduj abeyar lydp isd yyd yd saws me ydp yap yd yl sa1E15 jopz abeyqyme 13 10 glopajas pajepdn siweu g JOaz sapow ouuog Iduj ydp yap yd yr abeyy A EJO peqpasypaads N o q 13 10 did MEA D 4 l MEAD LSD ESA LaBEYOAOL aBeyOAGL doo7 pasojg Z doo7 vado Liopajas An yndup pyshor sopales U sudyn D BuO peag Hh SIXE PRAUIU 89 7 Control System Design 7 3 Speed Control The requirement for speed control is essential without it the diwheel is static and unin teresting Speed control can be achieved via open loop or closed loop control In the case of open loop control a voltage is applied to the armature of the motors by the motor controllers as described in Section 5 1 The basic effect is that the motor speed increases as the applied voltage is increased However a Permanent magnet DC motor cannot achieve equal speeds for all conditions For example if the load applied to the motor is increased the steady state speed will be reduced Hughes 2006 Closed loop control
29. a petrol powered diwheel by Jonas Bjorkholtz Di wheels 2009 2 1 1 Outer Wheel The outer wheel is used to make rolling contact with the surface of the road Similar to conventional bike and car tyres its function is to support the loads of the structure upon the road as well as provide some vibration and shock isolation from the variation in road surface Manufacturing a large outer wheel is problematic because the outer wheel must be capable of enduring contact pressures between the wheel and the road and between the outer wheel and the idler wheels Under operational loading local deformation must be kept to a minimum and the wheel s circular profile must be maintained Many diwheel manufacturers have constructed the outer wheel from tubular metal The wheel is typically metal tube or section rolled and joined to form a rigid ring The ROCKET ROADSTER made by Kerry Mclean shown in Figure 2 4 features a tubular aluminium wheel The wheel is rolled into a ring and then cut and welded It is then held flat heated and annealed until it is round Parks 2005 Figure 2 4 The Rocket Roadster by Kerry Mclean pictured Parks 2005 THE TRINITY constructed by Dave Southall Figure 2 5 uses a rolled and welded steel 2 Background wheel A tubular wheel allows the contact between the outer wheel and the ground to remain at a single point when the wheel is tilted In order to provide steering a monowheel s outer wheel must maintain contact
30. abdomen and shoulder straps down The restraint of the occupant also relates to how far the upper anchorage is mounted above the shoulders If these points are too far above the shoulders they provide little to no vertical restraint of the occupant Figure 4 7 Left Inverterd Y harness three point harness Middle Four point harness Right Five or six point harness Ottoson and Lovsund 1989 A racing harness combined with a regular seat can lead to submarining due to the lack of body support Ardoino 1984 Racing harnesses constrain the occupant well when used with a racing seat Further protection is added by using a racing seat through appropriate holes in the seat which allow the harness to better fit the human form These seats may be constructed using a tubular frame and panel board covered with cushioning Advanced versions of racing seats are made from fiberglass or a Carbon Kevlar composite These provide an advantage of weight reduction over the tubular framed racing seats Due to budget and feasibility constraints the cheaper go kart seats when compared to the racing seats were previously considered a suitable solution Dyer et al 2009 While a go kart seat was used in initial designs of the space frame due to the lack of upper back support it was later deemed inadequate A racing seat which provides upper back and side support was therefore purchased to be used in conjunction with a 6 point racing harness to provide safe and
31. accompanying disc brakes axles bearings and chain and sprocket drives for possible use Further specifications of the wheel include e Tyre Width 90mm e Measured outside diameter approximately 275mm e Rim diameter 6 5 inches e Maximum tyre pressure 250 kPa For the purpose of friction driving the outer wheel a pneumatic tyre is useful When the tyre is pressurised at maximum pressure the contact friction will be minimised while the rolling resistance will also be at a minimum The pressure may be lowered to achieve higher frictional force however an increase in rolling resistance will also occur Figure 4 10 shows a photograph of the drive wheels used in EDWARD The hub has an inset housing for a bearing The axle supports the inner runs of the bearings and passes through the hub The brake disc is attached to the hub by three 4mm bolts An analysis of these bolts for their adequacy of strength is given in Appendix B 2 Figure 4 10 The mini moto drive wheels chosen for use in EDWARD 4 6 Motors This section presents a brief overview of the motor selection although Dyer et al 2009 discuss the selection in more detail DC motors were chosen over AC motors because power is provided to the motors by the batteries AC motors may be used with batteries but require an inverter which increases the cost and complexity of the motor controller The major criteria for the choice of motors included 28 4 6 Motors e Cost e Performanc
32. adequate restraint during the operation of the diwheel A polyethylene car racing seat was chosen as an inexpensive alternative to more expensive tubular framed or composite seats The JAZ 400 Polyethylene racing seat used in the final construction of EDWARD is shown in Figure 4 8 The racing harness chosen for use in EDWARD was a 6 point VELO TRS Magnum racing harness shown in Figure 4 9 26 4 5 Drive wheel Figure 4 8 The JAZ 400 polyethylene car racing seat chosen for use in EDWARD Figure 4 9 The 6 point VELO TRS Magnum racing harness chosen for use in EDWARD 4 5 Drive wheel The method chosen to drive the outside wheel was a friction drive system This is required because the outer wheel is a rolled tubular structure The friction drive system requires an appropriately sized wheel with a high friction compliable surface in order to form rolling contact with the inner surface of the wheels and provide sufficient traction so that the drive wheel does not slip The drive wheel selected for this function was a mini moto rear wheel hub and tyre because it is appropriately sized and is an inexpensive off the shelf component The drive wheel is a cast aluminium hub with a pneumatic tube and tyre providing compliance with 27 4 Component Selection the wheel with a constant contact pressure This tyre is slick having no tread which provides a continuous friction surface Among these features the drive wheel was also chosen as it had
33. and the drive system block diagram models this by applying the torque produced by the motors to a Joint Actuator which actuates the drive wheel joint The motor torque is derived from an algebraic equation relating torque to speed and the applied armature voltage which is discussed in Section 7 3 1 Any inertial or damping properties associated with the chain drive transmission have been omitted from the block diagram model following the derivation of Section 13 The block diagram also contains both voltage and current limits used to simulate the limited supply voltage and current ratings on the diwheel During initial simulations when the dynamics were being verified against each other these limits were relaxed to oo 00 volts and amps respectively On the left hand side of Figure 7 4 a sensor block is connected to Base of the drive wheel joint The sensor measures the differential angular velocity of the outer wheels with respect to the inner frame Assuming the drive wheel does not slip along the inside of the outer wheels the angular velocity of the drive wheel is N times this measured value The only friction effects included in the block diagram is the viscous damping caused by the differential motion of the inner frame within the outer wheels This damping consists of friction due to the idler wheels as well damping in the transmission 83 7 Control System Design To Motor 7 Machine Environment E x dx d2x
34. augments the open loop system by providing negative feedback from a sensor For the case of closed loop speed control the sensed output is the average velocity of the two wheels 5 P1 pr which is inferred by the encoder measurements The closed loop system then forms the error between the desired speed and the measured speed and uses the error to increase or decrease the voltage applied to the left or right motors Closed loop speed control becomes necessary if the load on one wheel is dramatically different from the other 7 3 1 Open loop Speed Control As discussed above open loop speed control does not require feedback and simply uses the inherent properties of the two DC motors to control their speed Assuming the mechanical dynamics of the motor such as inertia and friction are small negligible in comparison to the inertia of the drive wheel outer wheels in fact they are lt 1 then behaviour of an armature controlled DC motor is summarised by the differential equations di L ke V ae T ki which describe the change in current and motor speed 0 due to a voltage V applied to the armature and the electrical torque 7 produced by the armature current If the torque produced by the motor is zero no load then the motor speed is directly proportional to the voltage 1 rl This relationship then provides the possibility of using an open loop system to control the diwheel s speed purely by altering the aver
35. been conducted for a number of vehicles including a bicycle Zheng and Anwar 2009 and an automobile Esmailzadeh et al 2003 However the most relevant research to the diwheel has been done by Baker et al 2005 and Tarakameh and 13 2 Background Shojaie Motion of these vehicles is achieved by the rotation of two individually driven wheels The small differences in the velocity of each wheel causes an undesirable drift whilst the desired trajectory is a straight line This effect can be minimised by implementing a closed loop yaw controller The effective implementation of yaw control is achieved by calculating the dynamics of the system determining its open loop response and designing a closed loop controller to improve this response Previous research on the dynamics of differentially driven robots has been conducted by Tarakameh and Shojaiewho used a Newton Euler method to obtain the dynamics for a differ entially driven system However these dynamics did not consider lateral slip The dynamics of a Segway type vehicle were calculated by Baker et al 2005 who produced a detailed steady state model Yaw control may be achieved either by open loop or closed loop Because open loop systems do not employ the use of feedback they suffer from inaccuracies induced by steady state errors as shown by Cooney et al 2004 Their mobile platform with four Mecanum wheels was an autonomous vehicle This meant that whilst in operational no user in
36. chosen a Figure 6 30 Final design Chain drive Idler Guards A plant risk assessment is a compulsory component required by OHS and other legal docu ments Through the risk analysis and assessment the idler wheels were found to be potentially dangerous to human hands Incidents may occur under certain circumstances if a person makes contact with the inner side of outer wheel when the vehicle is being driven or being manually moved Since the sharp edged Nylon idlers are pushed towards the outer wheel by the spring dampers with high force magnitude hand injury due to idler run over could be potentially irreversible Therefore it became essential to enclose the idlers in protective guards The design requirement for the idler guards was to eliminate the possibility of making hand contact with the idlers This could be achieved by either warning the person in advance of grabbing the wheels or setting up physical barriers which block the access paths to idlers However a human s response time in this case is not fast enough for the person to remove their hand before it reaches the idler Therefore the latter method was chosen for idler guard design Figures 6 31 and 6 32 show the design of the idler guard The idler is fully enclosed in a profiled 3mm thick aluminum shield and the access path the gap between the guard and the outer wheel to idler has been sealed by folded industry brushes 71 6 Mechanical Design Figure
37. control strategies to actively reduce the slosh of the inner frame has never before been demonstrated EDWARD has proven to be an exciting vehicle which showcases a strong mechanical engineering component and will be used to demonstrate automatic control at the University of Adelaide The following report details the methods and efforts taken to achieve the construction of ED WARD Initially a background to the project is presented in Chapter 2 where the mechanical design and control of previous diwheels and monowheels is reviewed Chapter 3 specifies the goals of the project with regard to functional requirements and performance measures The mechanical and electric components purchased for use in EDWARD are described in Chapter 4 while the electrical hardware designed to support the implementation of control on diwheel is described in Chapter 5 Chapter 6 discusses the mechanical aspects of the diwheel and the concepts developed throughout the design process Chapter 7 contains the design of the control strategies of slosh and inversion and presents the derivation of the uncoupled yaw and translation dynamics Chapter 8 describes the design of the software implemented on the MCU which achieves both user and active control of the diwheel The results achieved during the testing of the diwheel and it subsystems are presented in Chapter 9 An analysis of the final design of EDWARD in the context of the project goals is provided in Chapter 10 Chapter 11
38. control system 107 8 Software Design Data Logging and Processing Consists of SDCARD ANGLE SCITOMATLAB AND FRTYPE SDCARD provides functions for writing and reading to the SD card while SCITOMATLAB has been used predominantly for periodic data logging using a Real Time Interrupt ANGLE contains code to implement the Kalman Filter described in Section 8 5 1 and FRTYPE contains code to implement a fractional fixed point data type as well as a number of other useful macros Control e STEERINGCONTROL Implements open loop speed and steering control and contains unimplemented code for closed loop steering control e PID This file contains a number of control structures for use in future Unfortunately due to time constraints these structures could not be used to their full potential and the control implementation remains to be streamlined in future 8 2 Input and Output data This section briefly describes the connections required for inputs and outputs on the system Joystick The joystick consists of two separate input systems the axis potentiometers X Y and the inputs which are connected to a circuit within the joystick that determines which button is pressed The X axis red and Y axis yellow are connected to PADO2 and PADOS respectively The following list presents the connection of the remaining joystick wires e Purple PHO e Orange PH1 e Brown PH4 e White PH5 e Blue PH6 e Yellow P
39. deceleration and steering are controlled by a joystick which is read by the microprocessor Slosh control has been implemented in EDWARD to reduce the intrinsic rocking motion during harsh acceleration and deceleration and the success and viability of this control is detailed Inversion control which aims to invert the driver and inner frame upside down and maintain this unstable position has been modelled and shown to be feasible for application on the physical system Implementation of inversion control in the future is proposed and would allow the driver to drive the diwheel in its inverted state At the time of this report the authors are unaware of any other fully operational electric diwheels and no diwheels implementing any form of active control exist to date This has created the opportunity to explore new grounds in the field of active control as applied to non linear vehicles by providing a fun fully functional and visually impressive prototype The group have achieved all core project goals as well as a number of extension goals 111 Executive Summary lv Disclaimer This report is the work of the five authors named below Any information that has been obtained from other authors has been respectively cited where used Christopher DYER Kane FULTON Jonathon HARVEY Evan SCHUMANN Tao ZHU Disclaimer vi Acknowledgements Firstly we would like to acknowledge Associate Professor Benjamin Cazzolato
40. distribution board e Convert stored electrical energy into rotational motion Electric motor Convert motor rotation to translational velocity x Transmission Drive wheel x Outer wheel e Brake the drive wheel Hand lever Disk brake calliper cables Limit switch to indicate braking to the processor e Provide the option of slosh control to driver 17 3 Design Specification Measure slosh rate of the inner frame Gyrosensor x Accelerometer Control slosh rate of the inner frame Processor Slosh controller Motor controller e Provide yaw control to the driver Measure yaw rate of system Encoder Control yaw rate of system Processor Control circuit Motor controller e Dynamic Braking Determine when to activate dynamic braking Joystick Activate dynamic braking Processor Dynamic braking circuit Motor controller e Measure battery voltage Resistive voltage divider Processor e Display battery level Processor 3 3 Technical Performance Measures The technical performance measures are used to determine whether the functional requirements have been met and are given in Table 3 1 18 3 3 Technical Performance Measures Table 3 1 Technical Performance Measures E Minimum What is Measured Requirement Open loop yaw control Differential drive Encoders Yaw rate of the diwheel 1 5 rad sec Brake Braking time from full speed Mecha
41. does not allow the level of control of a fuzzy controller For the balancing controller to capture the inner frame at the inverted position the inner frame must not be rotating too fast as it reaches the inverted position Inverting and multiplying the slosh controller gain does not provide as high a level of control as the fuzzy controller which remains the preferred method of swing up control Closed loop step response for 6 Kp 10 5 T T T T T T T T T T T S o 0 1 2 3 4 5 6 7 8 9 10 Time seconds Closed loop step response for 6 Kp 10 Y J ke g 3 5 J jo Time seconds Figure 9 6 Closed loop positive feedback controller 9 3 Software and Electrical Issues The initial motor controllers were developed based on an open source motor controller and constructed by the University of Adelaide electrical workshop However during the initial stages of testing several MOSFET transistors overheated and were consequently destroyed as shown in Figure 9 7 This was attributed to the PWM control of the motors which causes large currents to flow within the motor controller as discussed in Section 5 1 The motor controllers were replaced with an off the shelf motor controller that featured a dual high current H bridge controller This system required additional software design to be implemented on the Dragonboard before it was functional which caused minor delays The main issue with the current controller is that it i
42. flame C su C Steam C En C Laser beams LJ Is C Other issues L Ad C No LJ PPE xX NA Page 6 of 10 Project No 757 Issue No 1 Ss THE UNIVERSITY F ADELAIDE E Ei PROJECT RISK ASSESSMENT Hazard Identified Likelihood ture extremes Consequence Score Comments i e when and where hazard is present task activity Hierarchy of Control Current Controls Action Required YIN Tfr to CAR Y IN Can anyone be affected by tempera Exposure to high temperature Exposure to low temperature Other issues No NA k OOO Can anyone slip trip or fall due to OO WO OOO The location of the plant Uneven work surfaces Lack of safe guards eg rails Slippery work surfaces Other issues Inside frame rotation No NA Can anyone come into contact with Unlikely Fluids or Gases under high Major Medium Incidents may happen if belts are not applied properly pressure due to O El Su C En Is O Ad C PPE O El C Su En Is Ad O PPE Belts are inspected before every operation N NA C Failure of the plant O El C Nature of the plant O Su Other issues En C No O ls XX NA O Ad C PPE Page 7 of 10 Project No 757 Issue No 1 DS Mi EM PROJECT RISK ASSESSMENT Hazard Identified Likelihood Consequence Score Comments i e when and where Hierarchy of Current
43. for his indelible enthusiasm and support throughout the project We would also like to thank Philip Schmidt who was an integral part of the design of the system electronics and the source of many ideas Furthermore we would like to thank Silvio De leso Finally we would like to sincerely thank Bob Dyer who has assembled the physical system EDWARD with great skill and patience vii Acknowledgements viii Contents Executive Summary Disclaimer Acknowledgements 1 Introduction 2 Background 2 1 Previous Designs aba tele e a A a A a a ZTA Outer Wheel 4 ras ula E rd A A AA Ss Ln Tyre AI ER 241 3 Inner Wheel bs iaa Be wth Ser o a ata De ed hee AE a PALA Adler Wheel la 3 a a ela ade a Pe hd ee Doles DEV SE Eye ld eee lee a Pe ES eee ES A Das II AN II A A II OS Gd OPN nt 1 BG tel Sg 2 1 8 Seating and Harness hundo dan dad Oe Sey De Se Gears Boyne 2 1 9 CONCIUSION A tr O O A BONER a a a ee a s 2 2 System Dynamics and Research ooo a a ee Ds slosh Control ss aei a nd rt ea Re Se cats ee ed 2 341 Previous Work atado dr Serge BD es Nado ee do A he de 2 3 2 Relevance of Studies 0000000000000 eee 2 3 3 Application to the Diwheel n54 2 3 40 a i 2 4 Yaw CONTEO se el e A Bee Be Se a 3 Design Specification Sly te A a ee a Gerke A Mt aE ee Ae Na 3 2 Functional Requirements sn d erro ras da oa 3 3 Technical Performance Measures 4 Component Selection Mechanical Brake Selection
44. give 3 1 A 0 S A ap COS 0 7 bi2 0 9 arcos 0 bp az sin 0 cos 0 6 Ja sin 0 f 7 24 where D J Jz aR cos 0 7 25 and 1 gt a Ay oe al da ag COS 0 7 b 2 0 Jab1p Jarsin 0 6 aga sin 0 cos 0 7 26 7 1 3 Electrical Dynamics The mechanical dynamics presented assume a torque input into the system However as Permanent magnet DC electric motors have been used to power the diwheel the control input into the system is the voltage applied to the motors to produce a torque It has been assumed that the electrical inductance of the motors Lm is sufficiently small it may be neglected and therefore the current in the motor coil is an algebraic function of the supplied voltage V and motor speed m Nns 4 and is given by Rmi KmOm Vins 7 27 78 7 1 Analytical Derivation of the System Dynamics where Rm is the resistance of the motor armatures in parallel Km is the motor torque constant which is equal to the back EMF constant for SI units for both motors N is the ratio of the wheel radius to drive wheel radius and n is the drive ratio from the motor sprocket to drive wheel sprocket as a chain drive has been used to transmit power from the motors to the drive wheel The differential torque acting on the wheel and the inner frame generated by the motor in terms of the armature current is given by T ING hl 7 28 Combinin
45. inner frame The fidelity of the angle estimate given by the accelerometer is highest at low frequencies especially at dc but becomes useless at high frequencies Guerra 2008 For this reason a gyrosensor can be used to provide a better estimate of the inner frame angle at higher frequencies The gyrosensor outputs a reading proportional to the angular rate of the inner frame which may be integrated to estimate 6 T J gyrosensor dt 8 2 Ty The issue with the gyrosensor however is that it has a non zero bias error which generally increases with temperature When this bias is integrated for low frequency changes in 6 the estimate for 0 eventually increases in error until it becomes useless However a gyrosensor can perform well for high frequency changes in 0 especially if the bias error can be estimated and corrected for By combining the two sensors together good performance at low frequencies utilising the accelerometer and high frequencies using the gyrosensor can be achieved Several methods can be used to achieve this sensor fusion 112 8 5 Measuring Inner Frame Angle and Slosh Rate Figure 8 4 Use of the two axis accelerometer to measure pitch e Complementary Filter Combines the accelerometer estimate and gyrosensor integrated estimate through low pass and high pass filters respectively Generally slower than other techniques Cannot track gyrosensor bias Simple implementation e Ka
46. internal load caused by friction Because friction dictates that an angular offset is required for translation to occur it is desirable that the slosh control aims to reduce the slosh angular rate rather than the slosh angle The aim of slosh control is to improve the settling time of the slosh angular rate and reduce oscillations of the inner frame about the steady state slosh angle Not all three components of a standard PID controller were included in the slosh controller as some terms were deemed unnecessary The integral component of a PID controller is used to reduce the steady state error of the output which in the diwheel s case is the slosh angular rate As the inner frame reaches a steady state angle offset with a constant input voltage the angular rate will also converge to zero without the need for integral control The derivative component aims to reduce overshoot of the output in this case the angular rate and hence also the slosh angle and was considered for use in the slosh controller However utilising 96 7 5 Slosh Control purely proportional control of the angular rate showed reasonable results in simulations of the controllers in SIMULINK and the use of derivative control was dismissed in an attempt to keep the controller as simple as possible The use of a purely proportional controller for slosh suppression was also outlined by Well stead and Readman on their ball and hoop system The ball and hoop system exhibits dyn
47. is shown in Figure 2 11 of Ben Wilson s version of a monocycle Dez 2008 This design uses two thin wheels similar to roller blade wheels either side of a tubular inner wheel 2 1 5 Drive System The drive wheel system provides a means of transferring the torque of the motor or engine to the outer wheels In the design of a wheel with an axle coupling of the wheels to the driven 2 1 Previous Designs Figure 2 10 A cross section of left an idler wheel mechanism and right a drive wheel used to guide and drive the outer wheel respectively Vereycken 1947 Figure 2 11 Drive wheel and idler wheel detail of the Ben Wilson monocycle Dez 2008 axle is required However in the design of a centre less wheel the drive mechanism requires the transference from the drive shaft to the large outer wheel Components of this system may include drive wheels or cogs As part of the drive system the outer wheel acts as the internal gear annulus in a planetary gear system with the drive wheel as a single planet The planet rotates around a common axis of the planet and annulus when accelerating or decelerating One form of drive system that transmits power to the wheel is a toothed drive gear within the wheel which meshes with the outer wheel This design is featured in many patented monocycles seen on the Museum of Retro Technology website Monowheels 2009 Another form of drive is a traction between a drive wheel and the outer w
48. motor of at least 300 Watts was used Although Dyer et al 2009 describe several candidate motors the final motor chosen for the system is shown in Figure 4 11 These motors are rated at 900 Watts each and provide 2 2Nm of torque at 34 Amps The final motor was not geared so that the backlash in the transmission was minimized requiring the design of a transmission system Figure 4 11 Picture of the final motor monsterscoopterparts com 2009 29 4 Component Selection 4 7 Motor controller The motor controllers were originally not a purchased item instead being manufactured by the University of Adelaide s electrical workshop However during testing the fabricated controllers failed and the selection of a commercial off the shelf motor controller was mandated This controller is discussed in more detail in Section 5 1 Despite the fact that the initial motor controller was not chosen its electrical topology was In order to be capable of driving the motors forwards and backwards as well as applying braking torque in the forward and reverse directions a 4 quadrant controller was required Dyer et al 2009 describe an alternative to the final choice of a 4 quadrant H bridge which consisted of a controller using two contactors arranged as a double pole double throw switch Although this solution was a viable one it was decided that its bandwidth would be limited due to the slow switching of the mechanical relays and was not pursued fur
49. testing of the motors with the batteries one of the controllers failed because a number of the MOSFETS transistors overheated This was later investigated as being caused by PWM control of the motors which has the effect of causing higher currents to flow within the motor and hence through the motor controller due to the flyback effect of the motor s inductance Because the inadequacy of the initial motor controllers was identified at a late stage in the project a more robust motor controller was promptly sought after The controller shown in Figure 5 3 was removed from a previous project that was no longer in active use The controller is a AX2850HE ROBOTEQ dual H bridge controller which runs from 12 43V It is for this reason that the full 48V bus voltage provided by the four batteries could not be used The controller can be controlled via three methods Analog control standard R C control and serial control 38 5 1 Motor Controller Figure 5 2 Initial H bridge motor controller Synthesis of a R C compatible PWM signal was the simplest to program using the Dragonboard and was pursued first Clark et al 2005 describe several issues with the process of controlling the ROBOTEQ controller however none of these issues were faced during testing Use of R C control restricts the use of some of the optional inputs to the controller such as those which disable the internal power MOSFETS but has been completely adequate in providing control
50. that occurs as the vehicle is accelerated or decelerated when applying torque to the drive wheels This was deemed necessary after viewing videos of monowheels and diwheels in operation If implemented correctly it would also allow maximum deceleration of the vehicle if necessary which occurs when the centre of gravity of the inner frame is horizontally aligned to the centre of the outer wheels The project was subsequently called the Electric Diwheel With Active Rotation Damping and the vehicle was nicknamed EDWARD Another control mode was also considered with the aim to make the ride more ex citing This involved a swing up controller followed by a balancing controller together termed an inversion controller The purpose of this controller was to invert the driver and balance the inner frame enabling the driver to drive around upside down A literature review was undertaken to evaluate previous diwheel and monowheel designs from which the design of EDWARD was to improve Using these designs as a basis an original mechanical concept design of a diwheel was produced and subsequently constructed The outer wheels were rolled and welded stainless steel tube with a rubber strip bonded on the outer rolling surface An inner frame supports the driver who is held in place by a five point racing harness The inner frame runs on the outer wheels with three nylon idlers that are coupled to the inner frame by suspension arms which act to provide some suspensio
51. the completion of this goal Dynamically stabilised pitch The controller may be engaged to turn the user upside at standstill and maintain this unstable static equilibrium point against disturbances This goal also requires the fitting of a harness to the seating arrangement The control strategies to invert the occupant and create static and dynamic stability while inverted have been designed for the mathematical model using fuzzy control for swing up and a linear quadratic regulator for stabilisation Feedback sensor issues and time constraints have prevented this strategy from being implemented on the physical system A racing seat and harness have been implemented to restrain the passenger during operation and system proving has seen the diwheel perform full rotations of the inner frame while maintaining driver safety These rotations have been achieved through manually rocking the inner frame back and forth and by locking the mechanical brakes while traveling at full speed During slosh testing the slosh controller was changed to positive feedback with an increased gain causing the system to attempt to generate slosh rather than reduce it This caused the inner frame to complete a full 131 10 Final Design Analysis revolution and illustrates a form of swing up control however due to the absence of a balancing controller full inversion control has not been completed Design of an alternative tire for the wheel The alternate tire ma
52. this controller was to invert the driver then stabilise them enabling them to drive around upside down In this paper the dynamics of a generic diwheel using a Lagrangian formulation are derived The control laws for the two control strategies are presented Details of the EDWARD diwheel and its parameters are used in a numerical simulation for which the open loop response and various closed loop responses are presented Finally suggestions for future control strategies are made 2 Dynamics of the 2DOF system In the derivation of the diwheel dynamics that follows motion has been restricted to the xy plane In this two degree of freedom model the left and right wheel and left and right drive wheels are combined into a single degree of freedom In this way both pairs of wheels and drive wheels rotate at equal speeds so that the diwheel does not yaw about the y axis A Lagrangian approach has been used for the derivation of the dynamic model of the diwheel similar to that of shown in Martynenko and Formal skii 2005 The following assumptions have been made in the derivation of the dynamics e The motion is restricted to the zy plane e Friction is limited to viscous friction and the Coulomb friction arising from the idler rollers is ne glected e The suspension arms are fixed keeping the centre of gravity of the inner frame a fixed distance from the centre of the wheels e The inductance of the motor is negligible and ther
53. was constructed using SIMMECHANICS A virtual reality VRML model shown in Figure 5 was built to aid in the visualisation of the plant Figure 5 Rendered image of the VRML model The various control strategies detailed in Section 3 were also integrated in to the SIMULINK model and the simulation results are presented following 4 1 Open loop response Figure 6 shows open loop response of the inner frame angle 6 and control voltage Vm 0 when the inner frame was rotated to just less 180 and released The slosh rocking is clearly evident 4 2 Closed loop slosh control Figure 7 shows the inner frame angle 0 and control volt age V when the inner frame was rotated to just less 180 and released Note that saturation of the motors just occurs and the slosh is dramatically curbed 4 3 Swing up control The two swing up controllers were investigated for the diwheel when it was initially at rest and positioned at its stable equilibrium Positive velocity feedback Figure 8 shows the response of the simple positive ve locity feedback swing up controller Since it is based Open loop response 100 T 9 degrees Control Volts 50 4 Amplitude Figure 6 Open loop response when rotated to almost 180 and released Closed loop response 50 775 7 y 4 z Open A nm 50 9 degrees 3 Control Volts E lt 100 150F 200 L P
54. whether the design is suitable or not Reconnect broken parts with stronger welds or add reinforcement bars Select stronger springs or use more in parallel Potential Causes of Failure Gear bearing failure Gear shafts are not in right position Gear wear out Gearbox cover is loose and shafts slip out from bearing holes Function Failure Mode Effect of Failure Causes partial or total loss of mechnical power and maybe unable to hold inner Gears loss mesh frame in place while operating in up side down condition Generate noise DN Current Process Controls D CRIT Recommended Actions Tighten bolts and screws on gearbox Regular inspection Limit mechnical power in put so gears are not overloaded and slip Change gears if they worn out Tighten bolts and screws Replace damaged bearings Check if the failure is caused by design faults Wheel suspension is unable to provide enough contacting force for the driving wheel to transmit all motor power to outer wheel Driving wheel rubber doesn t provide enough friction force Keep wheel surfaces both outer wheel and drving wheel clean and free of lubricating oil The design actuating suspension force is more than enough to generate adequate friction force for motor Clean wheel contacting surfaces Replace suspension springs with stronger ones if necessary Coat inside of outer wheel with rubber if the above two didn t work Dr
55. with the ground at a single point However this is not required in the operation of a diwheel as the wheels are permanently upright The Swedish diwheel Figure 2 6 uses a rolled C channel section for its wheel which has a flat outer surface Figure 2 6 Diwheel by Swedish Jonas Bjorkholtz Di wheels 2009 The diameter of the outer wheel influences the dynamics of the diwheel system and its overall size and mass while constraining the space for the passenger and equipment The size of the wheel limits the use of existing wheel and tyre arrangements that are of conventional dimensions which would provide the system with a convenient pneumatic tyre The diameter of the wheel is large ranging from 1 96 metres in the Swedish diwheel Figure 2 6 to 0 914 metres the diameter of a mini monowheel by laFrance Bressen shown in Figure 2 7 The height and approximate wheel diameter of the Peugeot Moovie Figure 2 2 is 1 54 meters The Peugeot Moovie s body extends past the diameter of the wheels to house externally mounted spheres providing further stability and preventing tumbling 2 1 Previous Designs Figure 2 7 The mini monowheel of laFrance Bressen Wiley 1988 2 1 2 Tyre The tyre of a diwheel system is required to isolate the wheel from the roughness of the road and provide traction for propulsion Conventional tyres are typically made from rubber which provides excellent grip The Belgian patent Vereycken 1947 presented a cu
56. 00N This estimation takes into consideration the static weight of system 67 6 Mechanical Design components and one operator as well as dynamic loads due to non linear structure movements In the estimation a worst case load of 3000N was placed as a point load at the center of the 0 8m RHS the longer sections of the 6 frame sections in Figure 6 26 as this is the load Figure 6 26 Final design Triangular inner frame condition which results in the largest deformation The corresponding maximum deflections are calculated for a series of tabulated RHS Refer to SHS selection Appendix C For the load condition on this frame torsional rigidity is also an important aspect to consider since the frame needs to withstand torque from the four attached swing arms due to vehicle turning In this case relatively large side width RHS is preferred since the moment of inertia of RHS increases significantly with the side width Due to this reason a 50mm side width was selected Structural weight is another important issue The wall thickness of RHS should be as thin as it can practically be Therefore 1 6mm thickness was initially selected However all external attachments with the frame were to be connected by bolts which need to be supported by the thin walls of the RHS frame The wall thickness should also be chosen with consideration of bolt tear out It has been advised by the Adelaide University Workshop that from the manufacturing exp
57. 09 A computer joystick was chosen for the final user control as it is the simplest cheapest and most compact option A joystick has a small space envelope which is particular important for mouting on the inner space frame It comes with several generic buttons and provides two axis control in a design that is immediately familiar to users Computer joysticks are also very cheap provided only coarse control in comparison to industrial joysticks is adequate Purchase of an industrial joystick which are particularly robust is not an option for project EDWARD due to budget constraints The joystick chosen for the project is a Logitech Attack 3 shown in Figure 4 16 Figure 4 16 Logitech Attack 3 Computer Joystick Global Systems Solutions 2009 35 4 Component Selection Most computer joysticks are the USB variety and designed for plug n play This means they possess are a USB controller with flashed drivers For the project connection to the microcontroller via USB is not a simple process since the USB protocol is very complicated to implement The alternative is to use the actual sensors in the joystick which can be read as analog voltages into the Analog to digital converter ADC of the Dragonboard A two axis joystick generally contains two potentiometers attached to the stick in a gymbal arrangement Pushing the stick forward or pulling backward only changes the resistance of a y axis potentiometer for example while pulling le
58. 1 19m Therefore for the worst case yawing situation a width of 1 2m will prevent the diwheel from losing contact to the ground with one of its wheels This value was used as an initial estimate of the width of the diwheel The width is not restricted to be greater than 1 2m in order to prevent tipping the diwheel on it s side rather it is only to prevent the loss of traction between one of the wheels and the ground The width is also subject to space limitations and requirements B 4 Anthropometrics There exist few available anthropometric databases for civilians due to vast population sizes and the lack of measurement techniques Karwowski 2006 However there exist anthropometric data for some selected groups For example extensive databases are maintained for military populations An appropriate set of anthropometric data may be crucial for certain design situations especially where safety is a factor Therefore in order to select an appropriate database one must consider the possible user population and the data s application to the required design For the design of the diwheel a comprehensive database has been selected from NASA s Man Systems Integration Standards Christensen et al 1995 This database is free unlike the many ISO and military anthropometric data sets and comprehensive representing a wide population of male and female 40 year old persons Dimensions of the human form used from this document can be seen Figure B
59. 4 The dimensions used in the design of the space frame are given in Table B 1 with reference to the numbers labeling the dimensions defined in Figure B 4 160 B 4 Anthropometrics Figure B 4 Anthropometric dimensions and reference numbers for a male Christensen et al 1995 Table B 1 Anthropometric measurements of various dimensions shown above of a 40 year old male Christensen et al 1995 758 194 529 312 457 223 362 356 Sitting height Buttock knee length Knee height sitting Elbow rest height Hip breadth Chest breadth Foot length Foot breadth Mass 5th percentile cm in 88 9 35 0 56 8 22 4 52 6 20 7 21 1 8 3 32 7 12 9 29 7 11 7 25 4 10 0 9 0 3 6 65 8 kg 50th percentile 95th percentile cm in cm in 94 2 37 1 61 3 24 1 56 7 22 3 25 4 10 0 35 8 14 1 33 2 13 1 27 3 10 8 9 9 3 9 82 2 kg 99 5 39 2 65 8 25 9 60 9 24 0 29 7 11 7 39 0 15 4 36 7 14 4 29 3 11 5 10 7 4 2 98 5 kg 161 B Appendix 162 C FMEA Risk Analysis and SOP 163 Failure Mode Effects and Criticality Analysis for the Diwheel Project Diwheel Edward Supervisor Benjamin Cazzolato Symbols and abbreviation S Severity Rating from 1 no danger to 10 important O Occurrence Rating from 1 to 10 D Detection Rating This number represents the ability of planned tests and inspections at
60. 6 31 Final design Idler guards Figure 6 32 Final design Idler guard detail view 12 7 Control System Design Often the aim of control is to improve a system s dynamics by altering the open loop per formance until it meets a number of required design specifications However in the case of EDWARD a number of the proposed control strategies are not only envisaged to improve the performance of the system but also to provide a novel platform for control systems which has not been developed before The first principle control aim is to implement closed loop slosh control as discussed in Section 7 5 This control is used to suppress oscillations of the inner frame during normal operation It was noted during testing that the sloshing motion of the inner frame does indeed make the system harder to control making a slosh controller desirable for improving user comfort and the handling of the diwheel A second major control aim though initially proposed as an extension goal due to its difficulty is to implement an inversion controller which can swing the inner frame upside down and a balancing controller to keep it there Such a combination of controllers is described in Section 7 6 The remainder of the chapter presents the results of the two modelling routes taken to verify the diwheel s dynamics with two degrees of freedom 7 1 Analytical Derivation of the System Dynamics The dynamics are defined as the effect that forces and torques hav
61. Appendix Elliptical contact area a P Figure B 1 Geometry of two convex elastic bodies in contact Batchelor amp Stachowiak 2005 1 1 1 1 1 1 1 1 1 E Re Re Ra Ry 003 Doa 0 04 00 gt R 0 01756m The contact coefficient ko is given by 2 2 1 1 1 1 1 1 1 i k Y Rie Riy gt Ray 2 a Riy a Roy cos20 0 1 m To A 1 Rix Riy Roz R2y ae fats gt e ot a 2 ats aim ota gt i cos20 o 1 1 1 1 0 03 0 720 0 04 i oo ko 3 4722 ki ka k3 k4 and ks are then determined from kp using a set of graphs on page 300 of Engi neering Tribology Batchelor and Stachowiak 2005 The contact coefficients were determined to be k 1 1 k2 0 91 kz 2 0 k4 0 32 and ks 0 52 The radii of the elliptical contact a and b Figure B 1 are given by 3WR 3 x 1300 x 0 01746 ky 1 14 2 ON By i 4 71169 poy 3WR 3 x 1300 x 0 01746 ut 0 914 2 221 ica E i 4 711e9 156 B 2 Drive wheel Hub Bolt The maximum contact pressure at the center of the contact region is given by 3W 3 x 1300 max 1041MP P 2rab 2m x 0 002685 x 0 002221 a The maximum shear stress is then given by trek k4Pmaz 0 32 x 104 1 33 3M Pa B 2 Drive wheel Hub Bolt The holes for the three hub bolts to connect the disc brake to the hub of the chosen drive wheel is shown in Figure B 2 The torque applied to the drive wheel from t
62. Controls Action Tfr to hazard is present task activity Control Required CAR YIN YIN Can anyone injured due to Ergonomic issues due to C Repetitive body movement or O El posture O Su C Insufficient space O En C Excessive effort push pull O ls L Working at a height O Ad L Seating design C PPE C Poor lighting C Other issues C No XxX NA E Chemicals El C Radiation El Su C Fumes C En C Dusts EJ Is C Vibration C Ad E Noise C PPE C Toxic gases or vapours C Other issues O No XI NA Page 8 of 10 Project No 757 Issue No 1 DS Mi EM PROJECT RISK ASSESSMENT Hazard Identified Likelihood Consequence Score Comments i e when and where Hierarchy of Current Controls Action Tfr to hazard is present task activity Control Required CAR YIN YIN Does the plant generate significant environmental hazards C Energy consumption O El C Water consumption C Su C Hazardous waste C En C Hazardous emissions Is C Nuisance noise O Ad E Produce ignition to the PPE surrounding area C Other issues C No IX NA STEP 3 ACTION REQUIRED BY MANAGER SUPERVISOR AUTHORISED PERSON Note Sign off required by Head of School Branch on Page 1 L All action items have been transferred to the Corrective Actions Register RMSS x If no actions required and residual risk is medium to very high the activity and the hazard s have been transferre
63. Do not touch the electric motors after operation allow time for them to cool down OLDCODODOOOO ODODO Switch Off main switch of the vehicle before leaving it unattended Page 4 of 5 SOP No 1 Issue No 1 Hs SAFE OPERATING PROCEDURE ka EDWARD DIWHEEL Note This Safe Operating Procedure must be reviewed a after any accident incident or near miss b when training new staff c if adopted by new work group d if equipment substances or processes change or e within 1 year of date of issue Page 5 of 5 SOP No 1 Issue No 1 D Mechanical Drawings 183 Please notice that the hole center distance need to be maintained at 50 5 mm to fit the brake calliper Item 1 Quantity 1 This item is exactly the same as above one except that fitting hole columns swaped places refer to graph below Item 2 Quantity 1 DRAWN 05 27 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Brake calliper holder Steel TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 13 QUANTITY 1 for left and 1 for right Material all parts 8mm thick steel plate ITEM 1 Name Damper fitting plate top Quantity required 4 ITEM 2 Name Damper fitting plate sides Quantity required 8 ITEM 3 Name Damper fitting plate bottom Quantity required 4 05 26 09 CHECKED Ben Ca
64. ED SCALE 1 0 15 T QUANTITY 2 DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato This section is made of steel SHS square hollow section with 50mm side and 2mm wall thickness Quantity required is 2 DETAIL A 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Rotational inner frame member back TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 5 QUANTITY 2 Quantity 2 This section is made of steel SHS square hollow section with 50mm side and 2mm wall thickness DETAIL A RAWN 05 26 09 HECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Rotational inner frame member bottom TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 5 QUANTITY 2 Quantity required is 2 This section is made of steel SHS square hollow section with 50mm side and 2mm wall thickness RAWN 05 26 09 HECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Rotational inner frame member front TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 5 QUANTITY 2 Quantity 6 make this section with 50mm steel SHS square hollow section with 2mm wall thickness RAWN Charles 05 26 09 HECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR
65. EEL WHEELBIKE SUPERVISOR Ben Cazzolato TITLE Suspension bridge plate bottom SCALE 1 0 8 QUANTITY 2 Quantity 4 Weld both sides of both plates leave enough space between hole and base plate for M8 bolt head and nut on outer side of both plates O DRAWN 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 LENGTH 201 TITLE Suspension Bridge plate sides UNLESS OTHERWISE SPECIFIED SCALE 1 1 QUANTITY 4 need to be maintained parallel when welding Weld both side 4 inside conn to be right q finish is acca ers donot need ngled round petable This item will be subjected to heavy cyclic load All the joints need to be strong lt E Quantity 2 DRAWN 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Suspension bridge plate top Steel TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 8 QUANTITY 2 Shaded areas represent standard M12 threads RAWN 05 26 09 HECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Swing arm pin joint axial TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 1 QUANTITY 2
66. Goodarzi A and Vossoughi G 2003 Optimal yaw moment control law for improved vehicle handling Mechatronics 13 659 675 Gandhi P S Joshi K B and Ananthkrishnan N 2009 Design and development of a novel 2dof actuation slosh rig Journal of Dynamic Systems Measurement amp Control Vol 131 p 5 Genta G and Morello L 2009 The Automotive Chassis Volume 1 Components Design Springer New York Grasser F D Arrigo A Colombi S and Rufer A 2002 Joe A mobile inverted pendulum IEEE Transactions on Industrial Electronics 49 1 107 114 Guerra R d S 2008 Kalman filter tutorial for balancing robot Technical report Haskell R E and Hanna D M 2009 Learning By Example Using C Programming the DRAGON12 Plus Using Code Warrior LBE Books Hughes A 2006 Electric Motors and Drives Fundamentals Types and applications Elsevier Ltd Jantzen J 1998 Design of fuzzy controllers tech report no 98 e 864 design Technical report Technical University of Denmark Department of Automation Karwowski W 2006 International encyclopedia of ergonomics and human factors CRC Press Lauwers T Kantor G and Hollis R 2006 A dynamically stable single wheeled mobile robot with inverse mouse ball drive In IEEE Int Conference on Robotics and Automation pages 2884 2889 Orlando FL doi 10 1109 ROBOT 2006 1642139 Li Y Gu F Harris G Ball A Bennett N and Travis K
67. H7 The joystick axis data is read from an ADC while the buttons states are read from a function called readJoystickButton 108 8 3 Steering and Motor Control Brake Depressed This circuit connects to PH3 and brings the pin high when the brake lever is pulled Motor Controller Outputs At this stage the ROBOTEQ motor controllers are controlled by two PWM outputs connected to PP3 grey wire and PP7 purple wire 8 3 Steering and Motor Control Steering and speed control is currently achieved through open loop control using the joystick A flowchart of the function that implements this control is shown in Figure 8 1 This function is called at 50Hz by a timer interrupt 8 4 Measuring Motor Speed The speed of each motor is measured by the encoders as discussed in Section 5 7 The speed is calculated using a constant time elapsed method from Li et al 2005 This method utilizes multiple registers from the Enhanced Capture Timer including the Pulse Accumulators A PT7 and B PTO one for each encoder PT6 and PT1 as an input capture of the second pulse train of each encoder and timer 4 as an output compare timer providing sampling instances Each pulse counter is sampled in succession commencing simultaneously with the timer on the falling edge of the pulse train After a specified time has elapsed Tel timer 4 the code awaits the falling edge Upon the next falling edge the time elapsed and pulse count are recorded and the other pul
68. J2 Jiag 0 1t ar b12 A ar bia arb agar 0 J2 ar b 2 J2 ar b 2 J2b1 28 and 0 1 0 T E ET a 5 29 ar JJ Ji ar J2 ap The poles of this plant are at s 0 0 22 2 591 0 26 with the complex poles having a damp ing ratio of 0 083 The transfer function from 7 to 0 exhibits one zero on the origin which is expected as at the steady state 0 s 0 0 It is interesting to note that the transfer function from 7 to y exhibits two undamped complex zeros at s 3 441 The presence of lightly damped complex zeros is similar to that found in other systems exhibiting slosh such as the ball and hoop system Wellstead accessed Aug 2009 The implication is that if the motor is driven with a sinusoidal input at the frequency of the zeros then the wheel will stand still and only the inner cage moves Note that for the case of a voltage input u Vm then the damping term arising from the differential velocity of the frame and wheel increases from b12 gt b12 bm re sulting in open loop poles at s 0 0 34 2 551 0 37 and the state input matrix B needs to be multiplied by Nn sKm Rm Linearising about upright inverted position Linearising the non linear dynamics given by Equations 19 and 21 about the upright position 6 7 0 p 0 gives the linear state equations 1 aR Ji Jo 0 0 aR h 0 x 0 0 0 aR gt Ji Ja Sidg 0 far J1
69. L 0 5 10 15 20 Time Figure 7 Closed loop response of slosh controller when rotated to almost 180 and released on exponential growth arising from the unstable pole it takes a considerable time 16 seconds to be captured by the inversion controller even when starting from an initial state of 09 5 Fuzzy controller Figure 9 shows the response of the fuzzy swing up con troller It can be seen from the magnitude of the control signal in Figure 8 that the simple method is no where near as effective as a bang bang like fuzzy controller in Figure 9 for driving energy into the system 4 4 Inversion control Figure 10 shows the closed loop response of the frame angle 0 and the control signal Vm to an initial pose of o 170 when using the LQR given by Equation 37 Closed loop response 200 7 degrees Control Volts 100F 150 50F Amplitude Me a O 0 5 10 15 20 Time Figure 8 Closed loop response of positive velocity feed back swing up controller Fuzzy swing up controller 150 8 degrees Control Volts 100 50 Amplitude 100 150 200 7 0 Time Figure 9 Closed loop response of fuzzy swing up con troller The expensive control gains given by Equation 38 were analysed and although they were able to stabilise the plant the non aggressive control law was unable to cap ture the inner frame upon sw
70. LBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES sor LENGTH sod Oteo Drive swing arm part1 Steel UNLESS OTHERWISE SPECIFIED SCALE 1 0 7 QUANTITY 4 note that slots at the end of the parallel section are of different widths on two sides of the SHS As shown in detail A amp B Please cut slot through the two sides of SHS UETAIL B Weld around joints These sections are to be made with 35mm side 3mm thickness steel SHS please weld the joints Quantity 2 DRAWN 05 27 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Drive swing arm part 2 Steel TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 0 5 QUANTITY 2 Shaded areas represent Quantity 14 standard M10 threads DRAWN 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 LENGTH 201 TITLE Idler wheel and suspension axial UNLESS OTHERWISE SPECIFIED SCALE 1 15 QUANTITY 14 inner rotational Y frame member short inner rotataional frame member front inner rotataional frame member back ground joints Inner rotational frame memeber Y short inner rotataional frame member bottom inner rotataional frame member short Weld all sections together as shown above weld all
71. Results A A Btw O GER Ste MPR Se ee 85 7 2 6 Visual Simulation Using VRML 2 0 eur ee ee a A 85 7 2 7 Final VRML Model e 88 o Speed Controla dt ls es Y e a a de A eA Be 90 7 3 1 Open loop Speed Control as Dr A Ss de 90 TA Naw Controle Tes A EE A A te eR he Bd 91 7 4 1 Derivation of Yaw Dynamics Pale e IS E dE Ss 92 7 4 2 Open loop Yaw Step Response o a 94 7 4 3 Closed loop Yaw Response e o 94 2 OOS GOmbrOl acm 3 BS heeds dr a We a BAER E A eh 96 7 6 Inversion Control gt ca ono ioaad ae saoe s roa oeaio ra coos rau 99 7 6 1 Design of Fuzzy Swing up Controller ooo aaa a 100 T 6 2 Balancing Controller oe noas a ge GA A ee e a a a e o A a 102 7 6 3 Combining the Swing up and Balancing Controller 104 7 6 4 Controllability and Observability of the Linearised System 104 8 Software Design 107 8 1 Overview of Software and Design Methodology 107 8 2 Input and Output data 2 a ds Do ee OG ee e Be 108 8 3 Steering and Motor Control se Dar ea Vs 44 109 8 4 Measuring Motor Speed ua A ds A daa 109 8 5 Measuring Inner Frame Angle and Slosh Rate 109 8 5 1 Sensor Fusion using a Kalman Filter 114 8 5 2 Performance of the Kalman Filter 115 8 5 3 Arctangent Function 4 a0 RO BR BS BS ee 115 8 6 Implemention of the Control Systems oaa aa a a 116 3 6 L los Ah COROLA IA A A ta amp
72. THE UNIVERSITY OF ADELAIDE AUSTRALIA UNIVERSITY OF ADELAIDE FACULTY OF ENGINEERING COMPUTER amp MATHEMATICAL SCIENCES SCHOOL OF MECHANICAL ENGINEERING MECH ENG 4135 HONOURS PROJECT E D W A R D Electric Diwheel With Active Rotation Damping Authors Christopher DYER Kane FULTON Jonathon HARVEY Evan SCHUMANN Tao ZHU Supervisor Ben CAZZOLATO November 25 2009 11 Executive Summary This report details the design and construction of an Electric Diwheel With Active Rota tion Damping EDWARD as well as the implementation of active control methods and system proving A review of current and previous diwheel designs is conducted and the final design of EDWARD presents several novel improvements both to the mechanical design and the handling of the system Mechanical improvements feature an isolated inner frame utilising automotive spring dampers and the design is minimalistic clean and aesthetically pleasing Motion of the diwheel is achieved through the use of two DC electric motors providing a much faster response to commands than previous petrol systems The diwheel s two DC motors are driven through the use of a dual H bridge motor controller which receives a pulse width modulated signal from a microprocessor A number of integrated sensors including a gyrosensor accelerometer and two encoders are used to measure the behaviour of the diwheel and are used to implement ac tive control of the system Acceleration
73. a header An image of the gyrosensor selected is shown in Figure 4 14 Specifications of the gyro are given below e full scale range 300deg sec e operating voltage 4 75 5 25V e zero rate output nominal bias 2 5V e temperature measurement compensation for digital 33 4 Component Selection e single axis e digital serial port interface and analog output Figure 4 14 Gyrosensor Breakout board Sparkfun com 2009 4 11 Accelerometer In order to allow for the inverted pendulum mode of operation of the diwheel a measure of the angular position of the centre of gravity of the inner frame is needed This could be obtained by integrating the angular rate from the gyrosensor discussed in Section 8 5 However as the gyrosensor is prone to drift creating low frequency fluctuations another sensor is needed to obtain an accurate measure of angular position An accelerometer can be used as a tilt sensor measuring the angle at which it is rotated about one of its axes These values are commonly accurate at low frequency rates of rotation yet lose accuracy at higher frequencies Baker et al 2005 Combining the advantages of both accelerometer tilt sensor and gyrosensor provides a method of accurately determining the angular position over a broad range of frequencies The two sensors can be used in conjunction through the use of Kalman Filter as outlined in Section 8 5 The major parameter driving the selection of the accelerometer was
74. a steady state gain of 0 535 rad s V The effective maximum differential voltage is 14 4 Volts which is determined by taking the product of the maximum voltage K control weighting given to yawing motion and 2 because it is a differential system which gives 36 x 0 2 x 2 14 4 Volts This produces a maximum yaw rate of 7 71 radians per second under no load However this response has a two percent settling time of 16 1 seconds as shown in Figure 7 11 The two percent settling time is defined as the time it takes for the output to be constantly within a two percent error band of its final value This result can be improved by using a closed loop system with a compensator 7 4 3 Closed loop Yaw Response A closed loop system is usually only required for high accuracy applications as suggested by Santana et al 2002 However as yaw control is responsible for improving the safety of the driver reducing the two percent settling time is desirable This can achieved by using a com pensator which utilizes yaw rate feedback The yaw rate is first converted into a G force of the centripetal acceleration by 7 49 where mg is the mass of the driver and r is the distance between the driver s head and the center 94 7 4 Yaw Control yaw rate 7 2 rad s yaw rate 4 7 rad s Figure 7 12 Magnitude of G forces for yaw rates of the diwheel Ashcroft 2002 pg 235 states that if a person experiences a force gr
75. age voltage applied to each motor Transfer function between voltage and y at the lower equilibrium The dynamics of Section 13 assume the input into the system is a torque input which does not correctly resemble the behaviour of the DC motors which are voltage controlled devices Because the motor electrical dynamics are important and are not negligible they must be accounted for in the design Assuming the motor inductance is negligible with regard to the remaining system poles i e the electrical time constant is much lower than any mechanical 90 7 4 Yaw Control time constant the relationship between the motor voltage its angular speed and the torque produced is given algebraically by Vki KikeWm ma pea 7 35 Rm Rm where keis the back EMF constant of the motor The motor torque is increased by the transmission ratio n and then further increased by the drive ratio N r the ratio of the wheel radius to that of the drive wheel radius As described in Section 13 the motor speed is proportional to the differential angular velocity 6 of the outer wheels and the inner frame and proportionality is Wy Nns Nn 0 Therefore the coupled torque applied to the outer wheels and inner frame is Nn Augmenting the dynamics derived in 13 by including the electro dynamic torque of Equation 7 36 produces the transfer function of Equation 7 37 between the input voltage and average wheel speed p T Vki
76. ake Lever free Disable PWM Enable PWM Provide 100 battery voltage if X axis is 0 Otherwise if speed command is lt 50 Ky 50 If speed is gt 50 Ky 50 30 x Y axis 50 RequestSpeed Y axis x Ks RequestYaw X axis x Ky SLOSH CONTROL Slosh control enabled RequestSpeed slosh No RightCommand RequestSpeed RequestYaw LeftCommand RequestSpeed RequestYaw f Drive left and right motor with Multiply by batt pre patery gains commands via PWM Figure 8 1 Flowchart of the steering function used to drive the diwheel with open loop control 110 8 5 Measuring Inner Frame Angle and Slosh Rate Setup timer TC4 and i Pulse counters A amp B Sample Right encoder next time Sample Right encoder next time Start timer TCNT and pulse counter Encoder Right pulse or time elapsed no pulse Encoder Right pulse or time elapsed no pulse Has the timer elapsed TCNT TC4 2 Set new TC4 rae Set new TC4 Reset pulse Yes Reset pulse counter A counter B Which encoder is being sampled Is there a Left pulse Is there a Right pulse Has the timer elapsed again elapsed again Stop counter Stop counter record actual time record actual time Ta TCNT Ta TCNT No Left pulse No Right pulse Left Hz 0 Left Hz 0
77. amics very closely related to those of the diwheel although there is no translation of the ball and hoop system and the output is a desired hoop angular position rather than a desired hoop angular rate as is the case in the diwheel The slosh controller for the ball and hoop system used pro portional feedback of the slosh angle rather than the angular rate which is understandable as the system output is the hoop position where no slosh angle offset will be present at steady state To determine appropriate slosh controller gains a linearised transfer function between the motor voltage V and the slosh angular rate was utilised This transfer function was de rived from the linearised matrices about the lower equilibrium position and is given in transfer function form as 0 s 0 1249s 0 0105s 7 51 V s 83 1 0382 7 124s 2 478 The transfer function of Equation 7 51 was loaded into the SISO DESIGN GUI in MATLAB to provide a simple method of altering the controller gains and viewing the step response The root locus of this transfer function is shown Figure 7 14 where the closed loop pole locations are denoted by the pink squares A live updated step response of the transfer function was viewed and the proportional feedback gain was altered until a reasonable step response of the slosh angular rate was achieved The slosh controller proportional gain was chosen as K 31 7 52 giving the slosh angular rate step response in F
78. and the rolling friction between the outer wheels and the ground The damping terms were determined through physical testing by recording live data and using an oscillation decay method To determine the damping term associated with the sloshing of the inner frame b12 the following method was applied 1 The diwheel was set up with no power and seating a driver 2 The inner frame was rotated by some arbitrary angle and held by another group member 3 A video camera was started to record the sloshing behaviour 4 The inner frame was released and allowed to slosh down to the lower equilibrium position 5 Recording was ended and the footage sent to a computer 6 The initial angular offset was determined by measuring the angle between the centre of the top idler wheel and the vertical at the moment before release of the inner frame as shown in Figure 9 1a 7 The footage was then run until the inner frame reached the peak of its oscillation about the equilibrium position The new angle between the same idler and the vertical was measured and recorded as shown in Figure 9 1b The time of this peak was also recorded 119 9 Experimental Results and System Integration 8 Step 7 was repeated for each subsequent peak until the diwheel returned to its lower equilibrium position 9 The zero position angle was also recorded 10 The logarithmic decrement 6 was determined using 1 ze m TD as n Litn where n is the number o
79. applied to the drive wheels essentially acting as a disc brake 2 1 8 Seating and Harness Passenger integration presents a challenge in a diwheel where the dynamics and mechanics pose serious safety issues If the diwheel accidentally or deliberately is provoked to gerbil the passenger may remain upside down for sometime and must be properly restrained Previous models of diwheels and monowheels have not required such security and have relied on the centre of mass being far away from the centre of rotation in order to keep to occupant upright THE TRINITY Figure 2 5 uses a 5 point racing harness anchored on the frame in combination 10 2 2 System Dynamics and Research with a go kart seat Seats used in previous designs are varied including the use of motorcycle seats go kart seats and automotive seats Diwheel and monowheel seating has also been constructed from panelling integrated into the frame Figure 2 8 The seat and harness places significant influence on the human occupant During operation the forces on the occupant due to acceleration may cause damage if the occupant is unrestrained or in incorrect posture Therefore the use of a seat and harness combination must be considered with due care to occupant safety 2 1 9 Conclusion The mechanical components reviewed in this section give an overview of the previous designs and methods used to form a functional diwheel Each design must be considered in the context of its desired fun
80. ar b2 arbi 4g AR 0 J ar bi2 Ja ar b12 Job 30 and 0 1 0 ea a 31 ar JJ Ji ar J2 ap The poles of this plant are at s 0 0 25 2 27 3 07 Note that for the case of a voltage input u Vm then the damping term arising from the differential velocity of the frame and wheel increases from b12 gt b12 bm re sulting in open loop poles at s 0 0 34 2 12 3 40 and the state input matrix B needs to be multiplied by N p as per the downward linearisation case m 3 Control strategies In this section a number of different control strategies are presented for the two dimensional diwheel model It should be noted that no literature to date has been pub lished on control laws for either monowheels or diwheels This is not surprising given that previous diwheels and most monowheels were human or IC engine driven which are not amenable to automatic control the latter having dynamics with similar time constants to the plant The parameters used for the model and thus the con troller designs are detailed in Table 1 Most parameters were estimated from the solid model of the diwheel and rider with the exception of the damping terms which were measured 3 1 Slosh control The purpose of the slosh controller is to minimise the amount of rocking sloshing that the driver experiences when rapidly accelerating or decelerating This has par allels with slosh control in liq
81. around the joints if possible please make sure holes are in the exact positions shown above Quantity of the frame assembly 2 In Inner rotational In In DRAWN 05 26 09 CHECKED Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED ner rotational fra ner rotational fra ner rotational fra Quantity e memeber short e member front e member bottom rame member back DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato TITLE Inner ratational frame assembly 1 SCALE 5 1 QUANTITY 2 DETAIL L Inner rotational frame Damper fitting sides Damper fitting sides damper fitting bottom Please note the 5 mm and 10mm clearances in detail AB amp C so that welds of damper fittings do not intersect mainframe welds DETAIL Damper fittings in detail A amp B are item 2 damper fitting sides DETAIL A DETAIL B All the paired fitting plates need to be maintained parallel please control the weld distortion Overview Inner rotational frame All damper fittings are to be welded onto the main frame Weld both sides fo each damper fitting _ plate b N Damper fitting EN is identical for left an right side fittings s o DRAWN 05 26 09 CHECKED Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 20 LENGTH 201 TITLE Inner rotational frame ASM 2 UNLESS OTHERWISE SPECIFI
82. assenger mounted control systems electronics and batteries The space frame must be designed with consideration for a number of functional requirements A functional design tree showing the hierarchy of the design components for consideration is shown in Figure 6 3 The dynamics of the system determines the required mass distribution of the diwheel as well as the functional stability of the frame The ergonomics provided to the driver affect component spacing and sizing around which the design of the supporting frame is based The loading requirements of the frame are derived from the passenger and battery loading as well as the system s self weight and dynamic loading Furthermore the loading on the system is dynamic because of the ability of the inner frame to rotate Budgeting constraints will also have a large influence on the design of the frame requiring the frame to be easily manufactured using inexpensive parts The designs presented in the following sections culminate in the final assembly diagram of the space frame shown in Figure 6 4 6 1 1 Dynamics The distance of the center of gravity of the diwheel from its center of rotation affects the manner in which the vehicle will roll The project goals will depend on the ease with which the inner frame is capable of rotating within the outer wheels The required dynamics are a compromise between the ability of the system to rotate and the stability required for effective slosh control To ini
83. ated minimum braking distances vs travelling speeds is given in the following graph Min Brake distance vs Speed Minimum required braking distance m Travel speed m s General Safety Visual inspection of plant prior to use Unsafe plant to be tagged out and reported to Project Supervisor Keep all parts of your body and attire safely clear of the rotating and moving parts at all times Ensure scheduled maintenance for this machine has been carried out including scheduled testing of Emergency Stop Only authorised and suitably qualified and competent persons to operate this machine Guards must be correctly fitted to the machine at all times during operation DO NOT attempt to open or remove guards while machine is being operated Assistance from other project members to be sought when mounting and dismounting the vehicle NEVER LEAVE EDWARD DI_WHEEL RUNNING WHILST UNATTENDED Safety glasses must be worn at all times during the operation of this machine Closed Toe Type Shoes must be worn during the operation of this machine Loose hair to be securely tied back loose clothing to be rolled up and or secured loose jewellery to be removed Hearing protection to be worn where appropriate to the task being performed Leather Safety Gloves to be worn where appropriate to the task being performed and Ensure feet fingers and other body parts are kept clear of moving parts and from pinch points during operation
84. bly were available for purchase and no work shop labour was required Figure 6 22 Concept design 3 Parts list of warehouse caster assembly Design Evaluation of Concept Design 3 Concept design 3 has many major improvements It is not only more visually attractive but also better performing The design includes all the advantages developed from previous concepts and improves on or eliminates some of the defects identified earlier The installation of a suspension system makes the vehicle more comfortable to ride as well as providing a higher tolerance on the outer wheel rolling accuracy The spring dampers force the caster wheels to follow the outer wheel s profile closely and no slip condition is achieved Another advantage of concept design 3 is that all caster components are available to be purchased from local stores Therefore workshop usage and lead time are avoided The design has some problems which required improvement The friction force in the caster 65 6 Mechanical Design Figure 6 23 Concept design 3 Off the shelf caster assembly Figure 6 24 CAD model of the roller assembly assembly sleeve systems will wear out the supporting bolts rapidly This is not desirable as the design lacks robustness In addition the center wheels in the roller wheel assembly the middle smaller sized wheels in between of the red wheels in Figure 6 24 which support a major part of the radial load alone the caster assembly shaft d
85. board is replete with a number of peripherals that have been choosen to demonstrate the capabilities of the 12 microprocessor A selection of processor capabilities is listed below e 16 bit processor with ability to perform 32 bit operations l http www batteryspace com 31 4 Component Selection Figure 4 12 Sealed Lead Acid battery source for use in the project Jaycar 2009 e 8 channel 8 bit PWM modules or 4 channel 16 bit PWM modules e Enhanced Capture Compare timer module allowing flexible timing e Up to 91 digital input output lines 3 Serial Peripheral Interfaces 2 Serial Communication Interfaces 2 8 channel ADCs with 8 or 10 bit resolution Assembly instructions for use with fuzzy control These features make the HCS12 microprocessor a versatile choice for control The Dragonboard has been designed predominantly to bring out a number of the processors features and add peripherals useful for debugging and prototyping such as e Liquid Crystal Display e LEDs e 7 segement display e Piezo electric Speaker e Keypad input This board has been extensively in past Honours projects because of its versatility and flexible programming interface 32 4 10 Gyrosensor Figure 4 13 The Dragonboard development board manufactured by Evbplus 4 10 Gyrosensor In order to provide the system with active slosh control sensory feedback is needed to measure the angular velocity of the inner frame There ar
86. cal 127 10 Final Design Analysis braking The two calipers are controlled by a single hand brake lever which operates two brake discs directly coupled to the two drive wheels as outlined in Section 4 2 In the event of loss of power to the motors as a result of some fault or the emergency stop button being pressed the mechanical brakes may be applied to bring the diwheel to a stop A safe stopping distance may be achieved through controlled use of the mechanical brake and care should be taken not brake so hard that the inner frame locks with the outer wheels In the event that locking occurs the diwheel will tend to tumble and may not stop within a safe stopping distance Ensuring that the driver is well informed of the optimal braking procedures sees the requirements of this goal being met 10 1 2 Electrical and Electronic Goals Open loop differential yaw control The rate of yaw shall be controlled by analog data sampled by the MCU from the steering sys tem The diwheel shall be capable of turning through user control with repeatable results and consistent operation The nature of the steering control shall be intuitive to the user Steering of the diwheel has been achieved through the use of a joystick as described in Section 8 3 A minimalistic approach was taken to the user control of the diwheel and both forward and reverse operation along with steering was implemented using the joystick The steering command signal is sent to the Drag
87. ccording to ratio of maximum radial and thrust load on the roller The 39 degree angle allows the resultant force of maximum radial and thrust load to be perpendicular to the Zone A idler surface Zone B has a curved profile to minimize the wedge effect and load concentration explained in Figure 6 18 on the idler s center spot The curved profile is tangential to both the Zone A idler surfaces on both sides of the roller Zone C has a steeper angle than Zone A The purpose of the steeper angle is to prevent derailing from occuring when the centripetal force on the vehicle is excessive C EROA Figure 6 29 Idler wheel profile Drive System To achieve the desired speed goal without significantly reducing starting torque a drive system is required to transmit power from the motor to drive wheel in a ratio of approximately 1 7 shown in Figure 6 30 For the required transmission ratio both chain and belt drives are 70 6 3 Outer and Inner Wheel Design effective The drive system backlash should be kept small in order to allow transmission of the motor torque to the drive wheel to be smooth This requirement narrows the belt drive selection to toothed belt drives which are significantly more expensive than a chain drive Due to the high cost of a toothed belt drive a chain drive has been chosen The sprockets have been purchased at C amp A COMPLETE DRIVE SOLUTIONS According to the supplier s stock sprockets with 73 and 11 teeth have been
88. ce is that the gravitational term acts to stabilise the diwheel compared to the in verted pendulum robots which are unstable Solution to the Differential Equations of the Mechanical System The system of differential equations may be solved in terms of 6 and p to give E 1 A a A 0 pl tancs r b28 4 arcos big a sin 0 cos 0 6 Jia sin o 19 where D J Jz ag cos 0 20 and 7 1 A mle oo Pp 5 H arcos 0 7 b0 9 Jabip Jzarsin 0 6 aray sin 8 cos 0 21 2 2 Fully Coupled Electro Mechanical System Electrical Dynamics Permanent magnet DC electric motors have been used to power the diwheel It has been assumed that the electrical inductance of the motors Lm is sufficiently small it may be neglected and therefore the current in the motor coil is an algebraic function of the supplied voltage Vm and motor speed m Nns p 0 and is given by Rmi Kmm Vin 22 where Rm is the resistance of the armature of both mo tors wired in parallel and equal to half the resistance of a single motor Km is the motor torque constant which is equal to the back EMF constant for SI units for each motor N a is the ratio of the wheel radius to drive wheel radius and n is the drive ratio from the motor sprocket to drive wheel sprocket when using a chain drive The differential torque acting on the wheel and the inner frame generated by the motor in terms of
89. ced by rotating the board and shaking it at 10 and 13 seconds The red plot is the estimate of 6 using the arctangent of the two accelerometer readings As can be seen several portions of the accelerometer estimate are noisy corresponding to vibration of the sensor board at those times The green and blue plots show reduced noise better estimate which is the principle aim of the Kalman Filter As the accelerometer measurement becomes noisy the filter automatically prefers data from the gyrosensor Between 15 and 17 seconds the board was rotated towards 180 degrees Because the two argument arctangent function atan2 has a range of 7 7 the angle returned by the function is discontinuous at T m For this reason the filter should not be used close to this region of the function or the zero datum should be adjusted by using modular arithmetic or physically orientating the sensor board differently 8 5 3 Arctangent Function The implementation of the two argument arctangent function on a microcontroller can be achieved in a variety of ways but it is generally undesirable to rely on compiled libraries since they dramatically increase code size are slow though accurate as they generally use a power 115 8 Software Design Estimate for 6 from arctan accy accx 5 4 oe 2 v 0 D a 0 5 10 15 20 Time Estimate for 0 from Kalman Filters 0 5 10 15 20 Time Figure 8 5 Performance of the Kalman Filter Top Estimates for 0 us
90. centre of the wheels from the parallel axis theorem and ar Rema and ag e g Mz are constants of convenience Euler Lagrange equations The dynamics are found from d OL OL ee a ag ap ten 9 r 5 d OL OL ie y 5 5 Z balp 7 16 where b12 is a viscous damping coefficient related to the relative velocities of the inner ring 0 and the outer wheel y b is the viscous damping constant associated with the wheel rolling and is surface dependent and 7 is a differential torque applied to both the inner ring and the outer wheel by the drive wheel motor assembly Evaluating the terms for 0 d OL e T J20 apr cos 0 ar p0 sin 0 of sin 0 a an 96 Evaluating the terms for y d OL r 42 F wish A a 5 ar sin 6 J p ar0 cos 0 OL e e Differential Equations Therefore the governing dif ferential equations of the diwheel are given by r J20 b19 4 ay sin ar cos 17 and T Je P b12 a b1P aR 6 sin 0 47 0 cos 6 18 It should be noted that the above differential equa tions are similar to the equations of motion derived for the monowheel by Martynenko and Formal skii 2005 Martynenko 2007 with the exception of the rolling re sistance term b It is also similar to that for the self balancing two wheel mobile robots Grasser et al 2002 Ruan and Cai 2009 and the ballbot Lauwers et al 2006 where the only differen
91. ch rolls freely inside a rotating hoop Wellstead This system is very closely related to the structure of the diwheel It has also been outlined in Dodge s 2000 revision of Abramson et al s 1996 paper Dynamic Behavior of Liquids in Moving Containers that slosh is well represented by a pendulum whose mass oscillates relative to the tank The structure of the diwheel resembles a combination of both the pendulum and ball and hoop systems making the literature mentioned useful in aiding with the derivation of slosh dynamics of the diwheel Dynamic slosh control was implemented and investigated for both the two degree of freedom 2DoF actuation slosh rig and the ball and hoop system Gandhi et al 2009 The ball and hoop system investigates the use of a simple proportional feedback controller feeding back the slosh angle Although this showed promising results in suppressing the slosh of the ball the best slosh suppression results required long hoop position settling times Wellstead and Readman The Slosh rig used a Proportional Integral Derivative PID controller integrated with another controller The aim of the 2DoF slosh rig was to reduce the forces on a body arising from liquid slosh To accomplish this the forces were fed back to the controller which then determined the desired trajectory of the actuators This trajectory was then controlled using a PID controller feeding back the position of the actuators Using this control structure the liq
92. chemes for an inverted pendulum and cart system IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY 16 1277 1288 141 References Tarakameh A and Shojaie K n Year unknown Modeling of a differential drive wheel mobile robot by newton euler method Technical report Iran University of Science and Technology Tilley J 2004 Understanding Solids the Science of Materials John Wiley and Sons Unknown 1935 Gyro wheel car Modern Mechaniz June edition 87 Vereycken E 1947 V hicule deux roues solidaires l une de l autre Two wheeled vehicle supportive of each other Wellstead P Year unknown Ball and hoop 1 Basics Control systems principles co uk Wellstead P and Readman M Year unknown Ball and hoop 2 Control and analysis Control systems principles co uk Wiley J 1988 Inside the wheel The Complete Guide to Monocycles Solipaz Pub Co Zheng B and Anwar S 2009 Yaw stability control of a steer by wire equipped vehicle via active front wheel steering Mechatronics 19 799 804 142 A Publications Arising from this Thesis In addition to the work done towards the Project Contract the group has also produced a journal publication for the Australian Conference on Robotics and Automation ACRA This paper is included in the following pages 143 Control of an electric diwheel B Cazzolato J Harvey C Dyer K Fulton E Schumann T Zhu and Z Prime University of Adelaide Australia be
93. city of the motor to make it go unstable giving a controller of the form a Om u Vm 0 0 10 40 x 10 4 0 1 33 where the gains were chosen such that they were as large as possible without severely saturating the motors Fuzzy Controller Fuzzy control has been applied to the problem of balanc ing the inverted pendulum first by Yamakawa 1989 Since then various fuzzy logic controllers have been ap plied to various aspects of balancing problems Chang et al 2007 have described fuzzy controllers for both the swing up and balancing control of a planetary train type pendulum Their swing up controller consists of a bang bang control scheme with little regard to the ve locity of the pendant link however this strategy is not immediately amenable to the diwheel system since such a strategy tends to fight with gravity Finally Marty nenko and Formal skii 2005 describe a fuzzy swing up controller which they have used to swing up a pendulum by emulating an energy based approach About the controller The fuzzy controller developed here uses two inputs from the physical system the an gular position of the inner ring with respect to earth 0 and the rate of the inner ring 0 The single output con sists of a voltage between 48V and 48V In contrast to previous fuzzy controllers applied to the inverted pen dulum this fuzzy controller does not aim to limit the horizontal travel of the diwheel and the two chosen in
94. code has been used in the past by Clark et al 2005 Baker et al 2005 to implement control on the HCS12 microcontroller found on the DRAGON12 PLUS They report several issues with the conversion of SIMULINK models to runnable code especially those arising from the order of block compilation For this reason a complete system coded purely in C has been created The software for the platform has been coded in FREESCALE S CODE WARRIOR IDE us ing as a basis the LEARNING BY EXAMPLE LBE stationery Haskell and Hanna 2009 The LBE stationery contains a number of functions useful to quickly activate and use many of the peripherals onboard the Dragonboard especially the LCD which has been heavily used for debugging and eventually as a basis for user feedback in the form of a menu system 8 1 Overview of Software and Design Methodology The software used in the project has employed modular design Functions variables and definitions are collected into a C source and header file with a name which describes their function in the system In many cases a number of these file pairs can be grouped together to form independent tasks required by the software The following briefly describes these groups Sensor Consists of ADC c and h GYRO and ACCEL files These files are used to sample the sen sors connected to the Analog to digital converters on the board The gyro and accel files are currently used to scale the raw sensor data for use in the
95. control The mechanical losses in the constructed diwheel wheel reduce it s ability to roll and so a greater speed will be needed to invert This value is not a final goal and must be compromised with design restrictions of geometry and dynamic requirements The dynamics driving the design of the width between wheels of the diwheel is a worse case yawing situation If one wheel locks and the diwheel yaws about this wheel then a centrifugal force created from this rotation must balance with the weight of the diwheel to keep both wheels on the ground The distance of the center of gravity from the center of rotation of the diwheel is estimated to be he 0 3m and the width of the Di wheel is Other parameters used in the following calculation are shown in Figure B 3 If we assume that e The translational velocity of the diwheel is v 10km hr 10 3 6m s 2 778m s e The total mass of the diwheel is mr 240kg e The weight of the diwheel is w mrg e Centrifugal force of the diwheel about the locked wheel is Fe The centripetal force around the locked wheel applied at the center of gravity is given by mo mu 240x 33 3704 A 1 2 l If we take the moments around the point A see Figure B 3 POES he 0 2Ma wx gt 5 159 B Appendix Locked wheel pivot Figure B 3 Front view of EDWARD in a locked braking situation Sw ho 0 2 1 2 Fl 7R 3704 15 I 4 2 Tai s 03 k ppx ctx 2 a gt
96. ct outcome and this remains the case for project EDWARD The design should be made in accordance with budget constraints For the project the budget is particularly tight In the design phase this is the primary consideration for the selection of candidate parts Because cost is directly related to component quality it is challenging to make the design functionally sound while minimizing cost The aim of the project is to successfully achieve the primary project goals and to accomplish as many extension goals as the budget allows 6 3 2 Concept Solutions To achieve the design aims mentioned in the previous section several concept designs have been made and assessed Improvements to the concepts have been made based on comments and advice from the project Supervisor as well as a number of manufacturing companies In the designs that follow only a single side of the diwheel system is shown as the other side is designed to be symmetrical Design Overview and Description of Concept Solution 1 Figures 6 13 and 6 14 show the initial concept for the rotating diwheel system In this design the outer wheel is made of a rolled industry I beam Four caster wheel assemblies Figure 6 14 travel on the inner flange of the I beam establishing support for further installations on the caster assemblies In Figure 6 14 the four caster rollers the four small sized rollers with black rubber tyre and gray rim are designed to take the inward radial load alon
97. ction and behaviour before use in the design of EDWARD Another factor in considering previous designs is that none have yet implemented an electronic feedback control system and that the EDWARD diwheel is the first of its kind 2 2 System Dynamics and Research The diwheel is comprised of two wheels which are connected to an inner wheel that is rigidly connected to a central space frame where the driver of the vehicle is seated Each of the two wheels may be independently driven by an appropriate drive system which applies torque to the wheels In the lower equilibrium position the combined mass of the inner frame and rider sits below the rotational axis of the outer wheels When a torque is applied to the wheels via the drive system the inner frame will be displaced by a reaction torque This continues until the weight of the displaced center of gravity counteracts the torque due to gravity Rotation of the center of gravity is generally considered unfavourably as harsh braking or acceleration robs the diwheel of its forward acceleration When the drive wheels are driven at different speeds the diwheel will rotate about a vertical axis This steering can be controlled by a differential drive scheme and is very practical If the center of gravity of the diwheel is rotated 180 degrees from the neutral position the characteristics of the system resemble that of an inverted pendulum The pivot point is then located at the central axis of the wheel
98. culation of elastic non linear contact problems is the Young s Modulus F and Poisson s ratio v The properties of Nylon 6 6 were E 2GPa and v 0 39 Tilley 2004 p 301 For this investigation the standard properties of steel were used as E 200GPa and v 0 3 The friction caused by movement between both surfaces was modeled using the friction coefficent between steel and nylon of u 0 5 The maximum loading to be simulated on a single idler was determined by considering the situation where the diwheel is inverted placing all load on one ilder wheel If the mass of the diwheel is overestimated at 250kg and is dispersed between two wheels and considering the pretension in the other spring dampers a conservative maximum loading on a single idler was estimated at 200kg 2000N These material properties and loadings were used in a model Figure 6 10 where the single idler and outer wheel contact were cut along two symmetry planes 59 6 Mechanical Design Load on outer whee pp Idler wheel Symmetry planes Idler wheel bearing support Figure 6 10 FEA idler wheel contact setup 6 2 2 Idler FEA Results and Conclusions The results extracted from the solved ANSYS model were compared and validated with the calculated results from Hertzian contact theory found in Appendix B 1 A solved model dis playing a contour plot of the shear stress is shown in Figure 6 11 The principle results from this model include e Maximum shear s
99. d a low pass filter The MCU will automatically compensate for the dropping battery voltage If the voltage drops enough to indicate depletion of the batteries the MCU may disable the diwheel 129 10 Final Design Analysis A battery compensation circuit consisting of two resistors forming a voltage divider has been implemented in the diwheel providing feedback to the MCU proportional to the battery voltage level as shown in Section 5 3 A low pass filter was not required for the battery monitoring circuit as the control loop samples this voltage level at a low enough rate that any high fre quency noise is not a significant issue The output to the motor controller is multiplied by a gain proportional to the battery voltage effectively compensating for the dropping battery voltage The circuit that has been integrated into the diwheel meets the requirements of this goal 10 1 3 Miscellaneous Goals Mathematical Model of System State space model A mathematical model of the system shall be developed A full mathematical model of the diwheel is presented in Chapter 7 and incorporates forward and reverse translational dynamics steering dynamics and slosh dynamics The mathematical model has been derived using both an analytical method and mechanical method to provide verification of the results The dynamics give system behaviour at a level of accuracy adequate for the design of the desired controllers and indicate the completion of this goal
100. d by different individuals must accommodate many various statures The space frame structure has several design features that interact with the occupant The sizing and location of these features requires careful con sideration in order to safely accommodate a variety of typical users Anthropometrics Anthropometrics is the measurement of humans from which statistical predictions on the vari ations in human proportions can be made This data is often used in designs that require human interaction For the design of the diwheel a comprehensive database has been selected from NASA s Man Systems Integration Standards Christensen et al 1995 Dimensions of the human form used from this document in the design of the space frame can be found in Appendix B 4 Figure 6 5 illustrates the use of anthropometric data in sizing the space frame Similarly a mannequin was used also used in the design process Figure 6 5 which had similar dimensions to the anthropometric data found in Appendix B 4 Figure 6 5 The use of anthropometric data left the lines drawn are of human dimensions and right a mannequin is positioned in the space frame 6 1 3 Final Design of the Space Frame The final concept of the space frame design has been refined from concept designs presented in the preliminary report Dyer et al 2009 The final design has a focus on passenger safety while being easily fabricated In the interest of occupant safety a Finite Element Analysi
101. d early in the design as it was thought that the wear of the brushes would become an issue after extended use so that the part would need replacing However this option may be returned to in a future design as the speed measurement is likely to be quite accurate Although resolvers provide adequate feedback they require oscillator circuits to produce the speed measurement Encoders provide a feedback in a digital form with a greater noise immu nity and do not require a complicated control circuit Encoders were found to be costly when purchased as integrated units The encoder units typically include bearings an encoder sensor and encoder disc and are designed to be mounted directly onto the motor shaft However such sensors were found to be too costly for the design By utilising the brake s disc a cost effective encoder sensor was made using a pair of photo detectors The two common forms of encoder sensors are optical and magnetic Magnetic encoder sensors were found to be unable to sense the holes in the brake s disc because the disc was made of steel Suitable optical sensors were sourced from the University of Adelaide These optical sensors were of low cost and were mounted on the rotating arm providing speed feedback by sensing the holes in the brake s disc A photograph of the optical encoder sensor is shown in Figure 5 11 Encoders work by producing two digital pulse trains which are 90 degrees out of phase In the design the puls
102. d into harness Only operate vehicle in open areas and ensure no unauthorised persons are within the operational area Safety harness belt is to be fastened at all times during operation Never operate the vehicle at speed exceeding 15 km h Do not press any buttons for which the assigned functions are unfamiliar DO NOT lock the brakes unless unavoidable in an Emergency situation Fast acceleration and deceleration is forbidden DO NOT press stop button for normal braking purposes DO NOT attempt to make contact with any moving parts The swing up function is only to be demonstrated by project staff once supervisor is satisfied it is safe to do so When stopping at fully stopped condition ask project staff to switch off main power before undoing the safety belt Page 3 of 5 SOP No 1 Issue No 1 OE SAFE OPERATING PROCEDURE pd EDWARD DIWHEEL Operation of Brake LJ Press brake button for normal braking purpose Mechanical brake is only to be used in Emergency conditions such as power failure or control malfunctioning All power supply will be cut by applying mechanical brake a LJ For both electric and mechanical braking systems maximum braking efficiency is achieved when the centre of gravity is raised to the horizontal direction where the seat back is parallel to the ground a Operator of Diwheel should be aware of the required braking distance at certain speeds A schematic estimation of calcul
103. d or unflat 3 Regular TEPE CLN on all gt injuries travelling surface weld connections Bolt connection Structure failure and possible 10 Excessive load or unflat 1 Use friction connection 1 failure injuries travelling surface principle puwedi leds failure and possible 10 Excessive load or unflat 3 Regular imapecuons on all gt injuries travelling surface weld connections 3 Use friction connection Excessive load or unflat incil ad rinciple or introduce travelling surface P P reinforcemet Load excedes bearing s Carefully select approperite design load or inapproprite bearing type for the job and install with correct method bearing type is used Load is excessive or sudden impact force such as running Regular inspections on welds onto a kerb with very high and fix existing cracks speed Spring stiffness is too low Insure springs are strong Springs not locked in place enough and firmly in place 10 24 30 10 30 10 24 30 20 RPN 10 24 10 48 20 Recommended Actions Retain wheel circularity Fix rubber coating Give enough time for glue to form adequate bond force Straighten over bent members or replace if necessary Reweld or use bolt connection instead Change to stronger bolts or introduce reinforcement bars Reweld or use bolt connection instead Change to stronger bolts or introduce reinforcement bars Replace damaged component and check
104. d the two linearised systems are controllable This means that using only the combined motor voltage input all system states can be transferred to a desired final state in a finite time however this time may be long due to damping Observability refers to the ability to estimate the states which are not directly measured by the sensors The minimum number of sensors to be used on the diwheel are the gyrosensor and accelerometer Using only these two sensors the output matrix C becomes 1000 130 va calculating the Observability matrix O C CA A cam 7 60 104 7 6 Inversion Control Control Mode E Open Loop Ground Control Voltage Multiport Switch States Selector1 Fuzzy Logic Controller lt 0 17463 Compare To 10 degrees Multiport Switch1 Fuzzy Theta Inversion Gains Constant Selector Figure 7 19 Fuzzy swing up with Balancing Controller SIMULINK Model Fuzzy swingup with Balancing Controller Simulation 200 y T T T T T 150 100 50 6 degrees 50 100 Control volts A 505 5 10 15 20 25 30 35 40 Time Figure 7 20 Simulation of the Fuzzy swing up and LQR Balancing Controller 105 7 Control System Design gives 1 0 0 0 0 0 1 0 O 0 0 1 0 7 61 0 0 1 0851 1 1763 0 although the entire observability matrix is not given in Equation 7 61 the point to note is that the second column
105. d to the Hazard Register and communicated to the relevant personnel Page 9 of 10 Project No 757 Issue No 1 pma THE UNIVERSITY F ADELAIDE B2 ME PROJECT RISK ASSESSMENT RISK ASSESSMENT TABLES Likelihood Table CATEGORY DESCRIPTION Almost Certain Incident will occur at some time 0 1 month Likely Incident could occur at some time 1 month 1 year Possible Incident is possible to occur 1 year 2 years Unlikely Incident is possible but unlikely to occur 2 years 5 years Rare Cannot imagine that this could occur over 5 years Consequences Table CATEGORY DESCRIPTION Minor Effects unlikely to last until the next day First Aid Likely to affect employee the next day Major Medical Treatment injury needs formal medical treatment Critical Injury requiring extensive medical treatment and or hospitalization Catastrophic Injury resulting in death or permanent incapacity Risk Score Calculator Likelihood Consequences Almost certain Possible Minor First Aid Major Critical Catastrophic Unlikely Low Rare Low Risk Priority Table Descriptor Priority Action Very High 1 Immediate action required The activity should cease immediately and short term safety controls implemented Notify Manager and assess activity High 2 Implement short term safety controls imme
106. deter mines the control output in accordance with the membership functions and rules described in Section 7 6 1 The modular arithmetic and bias functions given in Figure 7 19 are required be cause the normal definition of 6 0 is at the stable equilibrium and not the inverted position Without performing this modular arithmetic the fuzzy controller would see a discontinuity in 0 around r r The modular arithmetic shifts the 0 zero datum for the fuzzy controller to the inverted position Figure 7 20 shows the performance of the fuzzy swing up and balancing controller simulated in SIMULINK using an estimate of system parameters As can be seen the swing up con troller swings the inner frame upside down in around 14 seconds at which point the balancing controller captures the inner frame and stabilises it 7 6 4 Controllability and Observability of the Linearised System Because there is only one input into the 2DoF the differential motor torque it is important to determine the controllability of the system s states Controllability refers to the ability of the control inputs to bring the system to any desired state in finite time It is determined by calculating C B AB A B A 1B 7 58 where the A and B matrices are the linearised state matrices given in Section 13 If Co has full rank then the system is completely controllable For both the inverted and stable linearised systems the controllability matrix has full rank an
107. diately Notify Manager and assess activity Medium 3 Short term safety controls implemented to minimise risk of injury Notify Manager and assess activity Corrective Actions within one month Low 4 Notify Manager and assess activity Corrective Actions within three months if possible Page 10 of 10 Project No 757 Issue No 1 Oa SAFE OPERATING PROCEDURE pd EDWARD DIWHEEL LOCATION DETAILS Operation of EDWARD Diwheel Date 10t October 2009 PREPARED B ame Position and signature ert names of the superviso R O and operato olved Name Position Signature Benjamin Cazzolato Supervisor amp Vehicle Operator Christopher Dyer Vehicle Operator Evan Schumann Vehicle Operator Jonathon Harvey Team Leader amp Vehicle Operator Kane Fulton Vehicle Operator Tao Zhu Vehicle Operator HAZARD IDENTIFICATION RISK ASSESSMENT See Risk Assessment dated 10 10 2009 HIGH SAFE OPERATING PROCEDURE DETAILS 1 YOU HAVE NOT COMPLETED THE COMPULSORY UNIVERSITY OF ADELAIDE OCCUPATIONAL HEALTH AND SAFETY INDUCTION COURSE AND 2 YOU ARE NOT UNDER THE DIRECT SUPERVISION OF A PROJECT SUPERVISOR Preparation operating area check LJ Area is free from grease oil debris and objects which can be tripped over LJ Area is clear of unauthorised people before commencing any operation Personal Attire amp Safety Equipment a Clothing must be tight fitting a Long hair must be worn up preferably usi
108. e e Integration with motor controller After selecting DC motors the choice was made between brushed and brushless Brushless DC motors do not require physical brushes to make contact with the commutator in order to drive the motor resulting in less friction and heat during operation Ellis 2000 This gives brushless motors the advantage over brushed motors in performance providing higher torque speed and power efficiency Ellis 2000 However driving a brushless motor is more complicated than a brushed motor as commutation must be achieved electronically Ellis 2000 This requires the use of a more complex motor controller which needs close matching to the motor Brushed DC motors are much simpler in design and can be therefore utilised with relatively simple controllers The main driver for the choice of motors is cost due to the project s limited budget and the increased complexity of both the motor and controller for brushless DC motors results in a much higher cost compared with the equivalent brushed DC motor It was deemed that the reduction in performance of the brushed motors in comparison to brushless motors was less significant in the application than the reduction in cost and hence brushed DC motors were chosen for use in the project Following the decision on the type of motor a motor suitable for the diwheel application was sourced Through past experiences and communications with motor providers it was recommended that a 24V
109. e fore the current is an algebraic function of voltage and motor speed e The rotational and translational inertia from the motors and drive wheels has been included in the inner frame e There is no slip between the drive wheels and the outer wheels e There is no slip between the outer wheels and the ground The model has three coordinates however the latter two are dependent e 0 rotation of the inner frame assembly about the z axis e pr Yr y rotation of the wheels about the z axis e x the displacement of the diwheel centre about an earth centred frame The right handed coordinate frame is located at the cen tre of rotation of the diwheel as shown in Figure 3 The positive x direction is to the right and positive y is down Clockwise rotations about the centre are considered pos itive The zero datum for the measurement of both the body angle 0 and wheel angle vy is coincident with the positive y axis Since the drive wheel is fixed to the inner frame body 2 the two masses may be lumped together However in the development that follows the energy associated with the rotational velocity of the drive wheel is omitted as it is considered negligible m J Figure 3 Schematic of a generic diwheel showing coor dinate systems mass distributions and states 2 1 Non linear dynamics The Euler Lagrange equations yield the dynamic model in terms of energy and are given by d OL OL F 1
110. e The motor current causes a measurable voltage drop across the resistor which is then fed into a differential amplifier Due to the high currents drawn by the motors these resistors become quite specialised and therefore expensive Another solution is to use a dedicated IC as described in Dyer et al 2009 42 5 5 Current Sensors Why current sensing Current sensing is useful in controlling the torque produced by the motors as this torque is proportional to current for permanent magnet DC motors Hughes 2006 However the motor torque can be estimated from the following formula ee ae Rar WwW a k k where e w is the motor speed e V is the armature voltage e k is the motor torque constant e Ra is the motor armature resistance e Tis the motor electrical torque This relationship indicates that torque 7 may be estimated from knowledge of the motor speed w and the voltage applied to the motor V This estimate is slightly problematic because it is determined partially by the load and partially by the duty cycle of the applied PWM The motor back emf can be estimated from the back emf constant ke of the motor but these measurements may not be very accurate especially since the motor speed is quantised and inaccurate at low speeds The alternative is to measure the current in the motor torque directly where T kT Usually this is achieved by placing a sense resistor either in series with the motor or in both low leg
111. e a major risk of failure which may result in human injury and severe structural damage This problem may be resolved by using Nyloc nuts However position of idler rollers may require adjustments due to structure deformation or roller wear This can not be done with Nyloc nuts since they are not removable The design also lacks a tensioning system on the idler wheel hence this would require the outer wheel to be rolled with great accuracy in order to avoid misalignment between the outer wheel and idler rollers However higher manufacturing accuracy dramatically increases the wheel rolling process cost In addition deformation due to impact loading resulting from a fluctuating road surface may also cause misalignment issues between the wheel and idlers Another problem is raised by the drive wheel tensioning system The tensioning design which is shown in Figure 6 19 does not includes a damping device to dissipate vibrational energy in the drive system This might influence motor operation or the accuracy of sensor feedback 63 6 Mechanical Design Overall this design possesses improvements from the previous concept design in terms of its manufacturing simplicity However a number of obvious problems may be catastrophic during operation In addition to the possible failure modes the design is not a visually attractive solution according to meritous sources Design Overview and Description of Concept Design 3 Figure 6 20 Overview o
112. e Implement a mechanical braking system Electrical Electronic goals e Achieve open loop differential yaw control e Achieve open loop speed control Implement a dynamic braking system e Accomplish closed loop slosh control e Implement adjustable acceleration and deceleration ramps Implement a battery compensation unit Miscellaneous goals e Build a mathematical model of system state space model e Create a virtual simulation of the diwheel Extension goals e Implement a regenerative braking system e Use state control of the pitch angle e Achieve closed loop yaw rate control e Build a method for dynamically stabilized pitch e Design an alternate tire for the wheel 16 3 2 Functional Requirements 3 2 Functional Requirements The functional requirements are requirements that must be met by the various systems on the project These requirements have sub requirements that must be achieved by the components The Functional requirements sub requirements and components are listed below e Provide traction with the road surface Outer wheel tire e Seating of the driver and support of the electronic components Main space frame Seat and harness e Suspension of the main space frame Inner ring assembly e Read user input commands X Y axis data from the Joystick Processor e Provide power to the system Batteries e Supply power to the system s components Power distribution board Signal
113. e change in states determined by the state matrix and the change in the covariance matrix e 0 1 dt Ok 1 dt gyro JLo loe Jl 1 dt 1 0 Bel 1 Bala a 8 8 where Pi is the covariance matrix which measures how well the sensors or model are esti mating the states and Q is the process noise covariance matrix which measures the covariance of the states due to modelling uncertainties Measure and update The measure and update step calculates the Optimal Kalman gain which minimises the process and measurement covariances providing a optimal estimate of the system s states The result of this step is an optimal estimate for the inner frame angle 4 and the gyrosensor bias dz P 4 3 Jr 5 e 8 9 ES E ema o ES 8 10 P 1 K 1 0 P 8 11 114 8 5 Measuring Inner Frame Angle and Slosh Rate 8 5 2 Performance of the Kalman Filter The process noise covariance matrix Q is a measure of the variance of the two states with respect to errors and uncertainties in the state model It is generally difficult to select appro priate values for the elements of Q especially if the states are not independent For application on the diwheel the Q matrix was tuned to the following with the assumption that the two states were independent var 6x 0 0 03 0 Q a var y 0 A ae it was noted during testing that increasing the magnitude of the diagonal terms of Q made the Kalman Filter trust the acceleromete
114. e illustrated in Fig ure 6 12 The stresses described in the following list are Von Mises stresses The results obtained were e Excessive stresses due to bending found around the seating area were ignored because of the point loading used in the model e Maximum stress in the tubing at the connecting brackets was found to be 45MPa 57 6 Mechanical Design e Maximum stress was found around the seat mounting rails of 65MPa The VON MISES stresses found can be directly compared with the yield stress of the material oy 250MPa Thus the following conclusions and improvements as a result of the FEA performed were made A safety factor of approximately 4 was achieved in the analysed design e An excessive load factor is sufficient in conpensating for overloading during extraneous operation of the diwheel Genta and Morello 2009 The framing supporting the seat was improved in order to not subject the space frame tubing to excessive bending loads e The final tubing used for the space frame was a heavier grade Onesteel s Tubeline CHS 26mm OD medium grade 2 6mm wall thickness AN Maximum stress around connecting bracket 45MPa 4 Excessive stress due to point loading Maximum Stress 65MPa Figure 6 12 Von Mises stress contour plot of the space frame 6 3 Outer and Inner Wheel Design The role of the outer and inner wheels is to provide rolling contact between the inner frame and the drive system of the diw
115. e on the motion of a system In the case of the diwheel the system dynamics determines the behaviour of each element of the diwheel as a result of torques supplied to the drive wheels from the motors The derivation of the diwheel s dynamics has been performed using two separate methods for the purpose of verifying the final solutions of each and to increase confidence in the accuracy of the results The two methods used were an analytical energy based approach using the Euler Lagrange method and a mechanical block diagram visual based approach using the SIMMECHANICS TOOLBOX in MATLAB 7 1 1 Mechanical Dynamics The analytical method for deriving the dynamics of the diwheel uses the Euler Lagrange equa tions This energy based method is relatively simple to apply to the diwheel and has been applied to other similar dynamic systems such as the Furuta inverted pendulum Astrom 2000 and a monocycle Martynenko and Formal skii 2005b In order to derive the dynamics of the system the bodies of interest of the diwheel must be defined In the interest of deriving the dynamics of the diwheel for the purpose of slosh and inversion control the motion of the system has been restricted to the zy plane This simplification can be made assuming the slosh and inversion occurs predominantly in this plane and that any yaw does not have a significant effect on the slosh inversion dynamics 73 7 Control System Design m J Figure 7 1 Schematic side
116. e several types of gyrosensors gyros that measure angular rate These include spinning gyrosensors laser gyrosensors and vibrating gyrosensors The most suitable type of gyrosensor to the project application is a vibration gyrosensor or MEMS gyrosensor These are small cheap and commonly available They provide very good accuracy and although most are influenced by temperature many also measure the temperature to provide compensation for its effects The gyro was sourced from SparkFun an online supplier commonly used in previous final year control projects The gyros have standard fullscale limits on the angular rate 50 150 300 and 500deg sec being common and the correct specification had to be matched to the project s application Choosing a gyro with a low limit of angular rate provides finer resolution with less sensitivity to noise If this limit is too low however the sensor may saturate and an important state of control is lost A gyrosensor with a 300deg sec limit was chosen to suit the expected range of angular rates of the diwheel In order to maximise options for optimising the measurement signal a gyro that provided both digital and analog outputs was chosen It was anticipated that an analog to digital conversion of the angular rate within the sensor itself would involve less noise and hence provide a more accurate signal The gyrosensor selected came mounted on a breakout board making connections to its pins simple through use of
117. e trains are produced when a hole passes between the photo detector The phase difference may be used to determine the direction of rotation and is achieved by placing one of the sensors slightly ahead of the other The frequency of either pulse train is proportional to the angular speed of the brake s disc The encoder sensor can also be used as position measurement sensor by counting the number of pulses and multiplying the number by the angular distance between the holes in the disk brake s rotor However because the control systems did not require an absolute measure of position this feedback has not been implemented The encoder sensors were mounted on a circuit board to allow the sensors to be correctly located so that the two pulse trains were 90 degrees out of phase The arrangement is shown in 5 12 A steel plate was constructed so the encoder sensors board could be located as close to the disc s holes without fouling the disc The plate was then screwed into the arm with four screws 46 5 7 Encoder Sensors Figure 5 11 Encoder sensors made from two infrared photo detectors Figure 5 12 Location of the encoder sensors on the brake disc AT 5 Electrical Hardware 48 6 Mechanical Design The final constructed prototype of the diwheel is pictured in Figure 6 1 The construction of the physical system was completed by the University of Adelaide s Mechanical Engineering Workshop with the manufacture of a number of pa
118. eater than 3G s they start to temporarily lose vision This effect is not desirable and in order to reduce the magnitude of centripetal acceleration on the driver the maximum yaw rate must be reduced Figure 7 12 highlights the G forces on the driver with the original maximum yaw rate 7 2 radians s and the proposed maximum yaw rate 4 7 radians s As shown the original yaw rate exceeds 3G s at 0 495m from the center meaning the driver would need to be enclosed in the circle formed by this radius In order to eliminate the need to keep the driver in this circle a new maximum yaw rate was selected 4 7 radians sec This rate produces a maximum G force of 1 351G s at 0 6m away which is low enough that the driver does not experience any undesirable effects Thus in order to produce a 4 7 radians per second response for a 14 4 Volt input a compensator with a proportional component was required M ma XTX q m3 X 9 81 al 7 49 The compensator selected for yaw control is a PD controller with low pass filter This compensator reduces the maximum yaw rate and also decreases the settling time The transfer equation of this compensator is found in Equation 7 50 The closed loop step response is depicted in Figure 7 13 which highlights a settling time of 6 01 seconds This closed loop settling time shows a 10 09 second reduction when compared to than the open loop settling time 3 8075 0 9065 Gils s 24s 41 7 50
119. ed to the relative velocities of the inner frame 0 and the outer wheel p b is the viscous damping constant associated with the wheel rolling and is surface dependent and 7 is a differential torque applied to both the inner frame and the outer wheel by the drive wheel motor assembly Evaluating the terms for 0 A 5 nb an cos 0 ar sin 6 7 18 a sin a ar p 7 19 Evaluating the terms for y e 2 ap sin 0 6 J ar cos 0 7 20 2 a ai 7 21 77 7 Control System Design Differential Equations With each term evaluated the governing differential equations of the diwheel are given by rT Ja bio 0 9 dg sin ar cos A 7 22 and T J b 2 0 bg ar 0 sin 0 ar cos 0 7 23 It should be noted that the above differential equations are similar to the equations of motion derived for the monowheel by Martynenko and Formal skii 2005a Martynenko 2007 with the exception of both damping terms b and b 2 It is also similar to that for the self balancing two wheel mobile robots Grasser et al 2002 Ruan and Cai 2009 and the ballbot Lauwers et al 2006 the difference being that the gravitational term acts to stabilise the diwheel whereas in the inverted pendulum robots the gravitational term causes instability Solution to the Differential Equations of the Mechanical System The system of differential equations may be solved in terms of and to
120. eel Err R p mi 2 where my is the combined mass of both wheels Third the rotational energy of the inner frame Er L 9 where J2 is the moment of inertia of the inner frame about its CoG Lastly the translational energy of the inner frame CoG Ex P 2t 91202 ma o e cos 8 62 sin 0 2 ra ss E 0 31 where ma is the mass of the inner frame and N is the ratio of outer wheel radius R to drive wheel radius r Thus the total kinetic energy of this system is Ey Ejr Eye Lor En ma Re e cos 0 6 e sin o 2 po PR E 11 2 2 2 Potential Energy The potential energy of the wheel is zero Therefore the total potential energy assuming zero potential energy at 0 0 is related to the change in height of the CoG of the inner frame and is given by Ep eg ma 1 cos 0 12 where g is the gravitational acceleration Lagrangian The Lagrangian for the diwheel is the difference in the kinetic and potential energies Ej Ep i a 2 2 p Ji R m L 5 2 2 J Reme cos 0 p0 2 0 e g ma e g M cos 0 13 This may be expressed compactly as J se iaz L zo tar cos 0 p sO 4 cos 9 1 14 where J J R m ma is the effective moment of inertia of the wheel and inner frame about the contact point with the ground Jo Ja e mg is the moment of inertia of the inner frame about the
121. eels to implement active control A conference paper detailing the derivation of the dynamics and control structures has also been submitted to the Austalian Conference on Robotics and Automation ACRA The presentation of EDWARD at the Uni versity of Adelaide School of Mechanical Engineering 15th Annual Honours Student Project Exhibition received the Caroma Dorf Industries Ltd award for Best General Mechanical Engi neering Project and has attracted significant media attention being featured on CHANNEL 10 NEWS and TOTALLY WILD This attests to the amount of work undertaken in the project by the students and its overall success 138 References 2005 Pedal powered di wheel concept vehicle Viewed 7 April 2009 2008 Dezeen design magazine 2009 Direct plastics nylon 6 rod Aboel Hassan A Arafa M and Nassef A 2009 Design and optimization of input shapers for liquid slosh suppression Journal of Sound amp Vibration Vol 320 pp 1 15 Ardoino P L 1984 Dynamics of the belted occupant implications for protection Tech nical report International Center for Transportation Studies Elsevier Science Publishers Amsterdam Astrom K K J and Furuta 2000 Swinging up a pendulum by energy control Automatica 36 2 287 295 doi 10 1016 S0005 1098 99 00140 5 Baker N Brown C Dowling D Modra J and Tootell D 2005 SON of EDGAR State space control of Electro Drive Gravity Aware Ride University of Ad
122. el using the states derived from either the analytical or SIMMECHANICS Figure 7 6 shows the result of this task 86 3 14158 om A on Delta Slider 0 0 1 Gaint Delta Axis of rotation of cage Manual Switch Rate Transition 7 2 Simmechanics Derivation of the System Dynamics Axis of rotation of outer wheel Bias for Gravity The angle definitions are as follows Phi is the rotation of the outer wheel relative to earth centred frame Delta is the relative angle between cage and outer wheel Theta is the rotation of the cage relative to earth Delta Theta Phi 0 557 cos 0 3022 sin 0 3022 1 2 Eccentricity of centre of rotation of drive wheel and centre of outer wheel a VRML Block Diagram b Image of the VR Model Figure 7 6 The VRML models Front01 position 261p04 rotation ZG1p01 translation InnerFramenrotatio RightDrivewneet ce RightDriveWheel rotati LefDiveneel center LeftDrivewheel rotation VR Sink 87 7 Control System Design Figure 7 7 Image of the final VRML simulation 7 2 7 Final VRML Model The initial VRML model was used only as a verification tool to ensure that the dynamics that had been derived for the two DoF model were correct The system s behaviour was visualised for each control mode in a single plane of motion To accomplish the core goal p
123. elaide Batchelor A and Stachowiak G 2005 Engineering Tribology Butterwort Heinemann UK Cardini S 2006 A history of the monocycle stability and control from inside the wheel IEEE Control Systems Magazine 26 22 26 Chang Y H Chang C W Yang C H and Tao C 2007 Swing up and balance control of planetary train type pendulum with fuzzy logic and energy compensation International Journal of Fuzzy Systems 9 2 87 94 Christensen J McBarron J and McConville J 1995 Nasa man systems integration stan dards msis Anthropometry and Biomechanics Vol 1 Sec 3 Clark M A Field J B McMahon S G and Philps P 2005 Edgar a self balancing scooter Technical report The University of Adelaide Cooney J Xu W and Bright G 2004 Visual dead reckoning for motion control of a mecanum wheeled mobile robot Mechatronics Vol 14 pp 623 637 Di wheels 2009 The museum of retro technology Viewed 14 March 2009 Dodge F T 2000 The new dynamic behavior of liquids in moving containers Technical report San Antonio Southwest Research Institute 139 References Dyer C Fulton K Harvey J Schumann E and Zhu C 2009 Electric Di Wheel with Automatic Rotation Damping Technical report The University of Adelaide Ellis G 2000 Control system design guide using your computer to understand and diagnose feedback controllers Academic Press Esmailzadeh E
124. els and diwheels where the centre of gravity of the inner frame is offset from the centre line of the wheels It is very prevalent as these platforms typically have low damping between the wheel and the frame in order to minimise power consumption during locomotion In addition during severe braking or accel eration the inner frame will tumble relative to the earth centred frame which obviously affects the ability of the driver to control the platform In March 2009 honours students in the School of Me chanical Engineering at the University of Adelaide com menced the design and build of an electric diwheel The vehicle was called EDWARD Electric DiWheel with Ac tive Rotation Damping DYER et al 2009 A rendered solid model of the platform is shown in Figure 1 and Figure 2 shows the current stage of construction of the platform The outer wheels are rolled and welded stain less steel tube with a rubber strip bonded on the outer rolling surface An inner frame supports the driver who is held in place by a five point racing harness The inner frame runs on the outer wheels with three nylon idlers and is coupled to the inner frame by suspension arms which act to provide some suspension and also main tain a constant contact force between the idlers and the wheel Two brushed DC motors each drive via sprock ets and a chain a small motor cycle drive wheel which contacts the inner radius of the outer wheel Thus the vehicle can be drive
125. ementation of a real time radio control Allowing testing to be conducted at a safe distance without a driver 137 11 Summary 11 2 Conclusion The aim of this project was to design and build an electric diwheel with active rotation damping This has been achieved with the completion of a physical system which has undergone system proving and rotation damping slosh reduction testing with successful results An extensive review of literature relevant to the project has provided inspiration for both concepts of the mechanical design and control methods to achieve slosh and inversion control The final physical model has complied with the high level of safety measures asscociated with the design in order to minimise the risk of any driver injuries Redundancy in the form of an emergency stop circuit has been added to give the driver sufficient safety measures to mitigate any issues that may occur during operation The ergonomics considered during the design phase have resulted in a reasonable level of comfort considering the space constraints of the physical system and the aesthetics of the design have also been well recieved by the public The mathematical and VRML model of the diwheel have proved effective in giving insight into the predicted behaviour of the phyical system and when compared to testing of the physical system have proved sufficiently accurate The derivation of the mathematical model allowed tuning of the slosh controller to be p
126. ent levels required to exert braking torque on the motor The user shall be able to enable or disable this feature If disabled the 128 10 1 Summary of Goals and Achievements braking current will be limited e g motor terminals open circuit When enabled dynamic braking shall produce a noticeable increase in deceleration when both motors are turned off The effect shall be designed to be comfortable for the driver The ROBOTEQ motor controller used in the final diwheel electronics is based on a H bridge configuration allowing 4 quadrant operation The regenerative current to the batteries during dynamic braking is limited by the controller to less than 100A to ensure the circuit breaker is not tripped Disabling the dynamic braking during operation has not been implemented as the ROBOTEQ controller automatically uses dynamic braking This is in contrast to initial designs incorporating open source motor controllers which gave greater control in enabling and dis abling the dynamic braking through use of the Dragonboard Should the power to the motors be cut via the emergency stop circuit dynamic braking will be disabled and only mechanical braking may be applied System proving has shown the dynamic braking to be significantly effective in stopping the diwheel quickly and comfortably Closed loop slosh control When this is disabled rapid acceleration deceleration of the diwheel shall cause the inner ring to gerbil and deflect by some amount
127. equired sign was determined using trial and error The gain was initially set as K 31 and testing showed no perceivable slosh control To ensure that the controller was in fact being implemented in software correctly the gain was increased by an order of magnitude to K 310 Testing the slosh controller with this gain resulted in the diwheel becoming unstable due to the non zero gyro bias with the inner frame being inverted in one and a half swings This positive feedback response can be seen in Figure 9 6 which shows that the inner frame angle exceeds 7 3 142 radians at around 2 5 seconds While this indicated that the sign of the feedback gain was incorrect it demonstrated the effectiveness of using positive feedback to swing the inner frame upside down The concept of this form of swing up control was considered early on in the project but was initially dismissed due to presumptions that the system damping would be too high and that the control signal would be too weak due to the sinusoidal slosh angular rate For this reason a fuzzy control swing up controller was kept as a primary swing up control method However it is not known at this time whether the positive feedback swing up controller was linear or whether it was swinging between full positive and full negative voltage in a form of bang bang control Although there is potential in this form of swing up control using positive feedback along 124 9 3 Software and Electrical Issues
128. equirement that must be satisfied To confirm that these functional requirements have been met there must be a way to assesses each component and this is provided in the form of a number of Technical Performance Measures EDWARD y y y a y Main Frame Drive system Idler wheels ro SRA Outer Tire Y Y Seat harness Friction Drive Y y Drive wheel DC Motor Mechanical Batteries Brake Transmission Motor Controller Power Amplifier Microcontroller Steering input Figure 3 1 System Hierarchy Diagram 3 1 Project Goals The project goals are objectives that are intended to be achieved in the final design These goals are divided into mechanical electrical electronic miscellaneous and extension goals All goals 15 3 Design Specification in the former three categories must be achieved in order to satisfy the project contract while the extension goals are optional It is for this reason that the extension goals have purposefully been posed as difficult to achieve and are not required for the normal operation of an electric diwheel The following is a summarised form of these goals Mechanical goals e Create a concept solution for an electrically driven diwheel e Build a diwheel to be driven by two DC motors including a frame to support one person and systems equipment
129. er bracket is shown in Figure 6 17 This roller design uses a V channel concept combined with a flat channel bottom The flat center part of the roller is designed to both share part of the vertical radial load as well as reducing the risk of initiating cracks due to the wedge effect explained in Figure 6 18 Slots on the brackets make the idler position adjustable to fit the outer wheel and allow slight deformation of it 61 6 Mechanical Design Figure 6 17 Concept design 2 Idler roller assembly exploded view This concept design has a new drive wheel tensioning system as shown in Figure 6 19 which better fits the design frame The tensioning system is made from a number of simple RHS sections to minimize cost Spring connections will need to be fitted on the right side labeled to provide a securing force on the wheel shaft Holes on the left side columns provide a pivot point for the drive wheel assembly to swing up and down 62 6 3 Outer and Inner Wheel Design Figure 6 18 Wedge effect caused by a load on the idler Figure 6 19 Concept design 2 Drive wheel tensioning mechanism Design Evaluation of Concept Design 2 This design has some fatal problems Firstly the idler roller assembly brackets are connected to frame gusset plate by bolts meaning that cyclic loading and vibration introduced by a rough road surface may loosen the bolts and cause the whole frame to fall due to idler roller derailment There is therefor
130. erformed allowing immediate and effective implementation on the physical system Furthermore the testing of different slosh controller gains on the diwheel has shown the gains tuned in simulation to correlate very strongly to the physical system Braking of the diwheel has been achieved both electrically and mechanically to increase user control Mechanical braking provides a method of deceleration in the event power to the motors is lost or cut Dynamic electrical braking is the primary form of deceleration and occurs automatically under normal operation as a result of upgrading to a commercial motor controller The motor controller used in the final design also provides regenerative braking during normal operation charging the batteries as the diwheel decelerates Limitation on the current that may be sent to the batteries during regenerative braking has been implemented through the use of a circuit breaker Inversion control although not a core objective of the project was attempted in order to further increase the novelty of the diwheel The design of this control was completed using the simulated VRML model but was not able to be fully implemented on the physical system due to time constraints caused by project delays Half of the inversion controller the swing up has been achieved using a positive feedback control law The outcome of the project has seen the accomplishment of all core goals and the construc tion of the first electric diwh
131. erience 2mm is the minimum wall thickness for this load condition Therefore the wall thickness of RHS has been changed to 2mm in the final design Suspension Design Figure 6 27 shows an overview of suspension structure design In the final suspension design the swing arm structure is used instead of the radial slip structure proposed in Concept Design 3 This change makes the design more durable and optimizes the damper performance since the dampers are designed to work in conjunction with a linkage system where the damper arm relationship is perpendicular On the suspension arms 8mm square plates are used to connect 68 6 3 Outer and Inner Wheel Design Figure 6 27 Final design Detail of the suspension configuration dampers to suspension arms and reinforce the geometrical stability of suspension bridges at the same time The suspension bridge beams are also formed by RHS with size 35x35mm The dimensions of the beams are calculated to ensure that all idlers are able to maintain contact with outer wheel so that derailing will not occur For ease of assembly and adjustment connections are through bolts Nyloc nuts are used to ensure they remain tight Bearing Selection Bearing are used on critical joints and axles as shown in Figure 6 28 In the system the bearings take mainly two types of load radial and thrust The radial load is due to the system s static and dynamic weight impact forces from road surface and pre tension f
132. ernal shroud However as discussed in Section 6 3 3 the risk of leaving them exposed was identified during the diwheel s construction The design discussed in Section 6 3 3 was the result of this realisation and caused delays in other project priorities by several weeks An initial prototype of the idler wheel guard was constructed from rigid sheet aluminium and contacted the outer wheel at various locations inducing a debilitating rolling resistance Subsequently the rigid guard was modified to provide clearance with the outer wheel This created a dangerous gap exposing the idler wheel and not solving the issue Flexible materials such as rubber hosing agricultural piping and nylon brushes were then considered to bridge this gap and to completely guard the idler wheels from human contact The nylon brushes succeeded in fulfilling the design requirements of the guard and were used without further deliberation 126 10 Final Design Analysis 10 1 Summary of Goals and Achievements The project goals were divided into Mechanical and Electrical Electronic goals at the initial stages of design The level of accomplishment of these goals at the time of this report is outlined here 10 1 1 Mechanical Goals Create a concept solution for an electrically driven diwheel A number of CAD designs will be created and reviewed The cheapest solution shall be chosen for further development Throughout the design process 3D CAD models of the diwheel we
133. ertaining to the VRML model the final VRML enviroment required a modelling of the diwheel s behaviour for all command inputs This encompassed the two DoF dynamics of the initial VRML model as well as the response of the system to a yaw command To achieve this goal with an aim to keep the system dynamics as simple as possible the yaw dynamics were considered as being uncoupled from the two DoF dynamics described in 13 This allowed the initial dynamics to remain the same while another set of dynamics was run in parallel to alter the heading and position of the diwheel in the virtual reality environment In order to view the results of the updated VRML model a joystick identical to that used on the physical diwheel was integrated to allow the user to drive the virtual diwheel in the same fashion as the physical system as well as giving the user opportunity to test each control mode without stopping the simulation To improve the visualisation of the speed and movement of the diwheel as well as allowing the option of a tracking camera view the environment in which the diwheel runs was updated Simple ground textures and trees were added to give indication of speed and yaw rate resulting in the VRML environment displayed in Figure 7 7 A view of the highest level mask of the final VRML model incorporating all the operating features of the physical system excluding mechanical braking is shown in Figure 7 8 88 7 2 Simmechanics Derivation of the System
134. f Concept design 3 The third concept design is shown in Figure 6 20 The major improvement achieved in concept design 3 is the integration of a suspension system on all idler wheel assemblies and the bottom drive wheel swing arm All components included in the proposed caster assemblies could be sourced from Bunnings Warehouse The frame remains to be made from RHS but connected in a triangle manner rather than the rectangular connection used in concept design 2 At this stage of the design the drive wheel brake disc and sprocket had been sourced from various suppliers and drawings of each component include in the design The caster wheel assemblies at the vertices of the triangular frame are free to slide in the outer wheel radial direction The constraint system of the caster assemblies contains a seriess of bolts and sleeves This is shown in detail in Figure 6 21 In Figure 6 21 the silver coloured tubes bolt sleeves run on the shank surface of the bolts The supporting bolts are lubricated to allow free rotation of the sleeves The slide system setup is similar to that of plain bearings but with a longer axial length The Roller wheel assembly has been designed from items available from a local hardware store More details of the caster assembly are included in Figures 6 22 6 23 6 24 64 6 3 Outer and Inner Wheel Design Figure 6 21 Concept design 3 Caster assembly slide path All the parts appearing in the figures of the caster assem
135. f peaks being measured across angles must be on the same side of the equilibrium level and x is the amplitude of the peak 11 The damping ratio was then found using p VAT 82 12 The average period T between peaks was found using the recorded times and was used to find the natural frequency wn by 27 Wn F 13 The corresponding damping term in the mathematical model presented in Section was then tuned until the dynamics revealed poles exhibiting this damping ratio and natural frequency To measure the damping term associated with the rolling friction between the outer wheels and the ground clamps were added on either side of the idlers effectively locking the inner frame to the outer wheels The testing process was conducted in the same manner to determine the second damping term The system parameters used in the mathematical model are given in Chapter 7 and simu lations have showed close matching of the sloshing behaviour between the model and physical system This suggests that system parameter identification methods have produced adequate results 9 2 System Tests To measure the success of the implementation of control on the physical system a test method was required to provide quantitative results of the diwheel s behaviour with and without con trol These tests were used to measure the level of accomplishment of the goals related to the implementation of control in the project In order to further verify the ma
136. from the neutral position Slosh control shall be implemented to damp this oscillation improving user comfort The success of this control shall be measured quantitatively and qualitatively through the use of a gyrosensor and driver reactions respectively System proving illustrated the significant sloshing of the inner frame during harsh accelera tions and decelerations The implementation of the slosh controller showed substantial slosh reduction in both the VRML model and the physical system as outlined in Section 7 5 and Chapter 9 respectively Logging of the slosh angular rate and feedback from driver reaction with slosh control implemented has shown that this goal has been achieved Adjustable acceleration and deceleration ramps When closed loop slosh control is disabled the user shall be able to customize the acceleration and deceleration ramp times by altering default parameters in the MCU Following the derivation of the diwheel system dynamics it was found that this goal is anal ogous to slosh control and allowing manual tuning of the slosh controller gain Therefore this goal and the closed loop slosh control goal are not mutually exclusive and a solution for one goal also solves the other Battery compensation Both open loop and closed loop control require a fixed gain across the amplifier A compensation circuit will be implemented to achieve a fixed gain regardless of battery voltage This circuit shall consist of the MCU an
137. ft or right changes the voltage of a x axis potentiometer The two potentiometers share a common ground and common voltage and are thus in parallel The signals from each potentiometer are derived from the respective wiper voltages according to the simple voltage divider formula The Logitech Attack 3 joystick comes with 11 buttons a rotary potentiometer and a two axis stick It is cheap sturdy and equally comfortable to both left and right handed users A discussion of the electrical connections to the Dragonboard can be found in Section 8 3 36 5 Electrical Hardware This chapter presents an overview of the electronic hardware used on the diwheel system The arrangement of the electrical hardware in the control box is shown in Figure 6 7 The electrical hardware consists of a number of components e Motor controller e Power board e Power control board e Ignition circuit e Current sensors located on the underside of the electronics box e Gyrosensor and accelerometer board sensor board e Encoder sensors Figure 5 1 Arrangement of electrical hardware in the control box 37 5 Electrical Hardware 5 1 Motor Controller The motor controller lies between the microcontroller and the motor itself It is required to apply the correct voltage and sink or source the required current in order to achieve the microcontroller s desired speed and torque commands The motor controller can be active or passive in the former case the mo
138. g Equations 7 27 and 7 28 gives the differential torque in terms of applied voltage T NngKim Vm NtisKm 9 Rm 7 29 Inserting Equation 7 29 into Equations 7 22 and 7 23 yields the differential equations of the fully coupled electro mechanical system By performing these substitutions equations 7 24 and 7 26 may be rewritten in terms of an input voltage to the motors to give 4 1 a EOS i arcos 0 Rm Vin 012 bm 6 9 a sin 0 cos 8 6 Ja sin 0 7 30 where bm WnsKm is the effective damping from the back EMF and Rm A e gt io Ja arcos 6 1 Nn Km a p ETEV b12 Ym 0 4 Jansin 0 anagsin 0 cos 0 7 31 7 1 4 Linearised Dynamics In order to be able to apply typical control tuning techniquies to the diwheel system a linearised model of the non linear dynamics was required The dynamics of the plant were linearised about two operating conditions the downward stable position and the upright unstable position as the control modes were required to operate about these positions Linearising about downward position Using a Jacobian linearisation the non linear dynamics given by Equations 7 24 and 7 26 about the downward position 0 0 y p 0 may be approximated by the linear state equations x Ax Bu 7 32 79 7 Control System Design T where x 0 p 0 isthe state vector u 7 is the plan
139. g MATLAB S Fuzzy Locic TOOL BOX This allowed the controller to be quickly designed and tested with minimum difficulty The first part of the design required the specification of the membership functions Because the objective of swinging up the inner frame does not require the diwheel as a whole to restrict its translational motion to limit y this did not feature as an input This is in contrast to Muskinja and Tovornik 2006 as well as others who have placed restrictions on this state because of the restricted travel on their platform Instead the fuzzy controller uses two of the system s states e 0 the position of the inner frame with respect to the gravity vector e the angular velocity slosh rate of the inner frame to produce a control signal As discussed in Section 7 6 the membership functions consist of functions which describe how much an input belongs to a number of input categories If the number of membership functions is large it is usually possible to produce a fine and relatively exact control signal but the complexity involved may become increasingly high Alternatively the fuzzy controller may be realised with fewer membership functions with a possible loss of performance such as poor robustness The input membership functions for 6 and were designed following the reasoning of Muskinja and Tovornik 2006 More membership functions were used for the angle of the inner frame than for the angular rate in order to p
140. g the outer wheel radial direction pointing towards the wheel s center of rotation and the long silver roller is designed to bear the outward radial load The two flanges on either side of the longer 59 6 Mechanical Design roller help maintain the position of the caster assemblies along the axial direction of the outer wheel Housing of all rolling elements is achieved by two bent metal plates with a 10 millimeter thickness Figure 6 13 Concept design 1 Overview Figure 6 14 Concept design 1 Caster exploded view The drive wheel tensioning mechanism envisaged for Concept 1 is shown in Figure 6 15 The tensioning mechanism was not developed past this stage and is only a representation of the tensioning system The design of the tensioning system is similar to those of the caster 60 6 3 Outer and Inner Wheel Design Figure 6 15 Concept design 1 Drive wheel tensioning mechanism assemblies In Figure 6 15 the longer roller with flanges represents the drive wheel with a motor connection shaft The four smaller rollers function in the exact same way as the caster assemblies The contact force between the drive wheel and I beam surface can be controlled by adjusting the tensioning bolts Advantages of this design are that this system is self locating and able to support itself and the motor assembly not shown in the figure Design Evaluation of Concept Design 1 In concept design 1 the outer wheel is manufactured f
141. gle The mathematical model of the system shall be used to provide state space control of the pitch angle State space control of the inner frame angle has not yet been achieved although the necessary controllers and control structures have been designed The implementation of these controllers remains to be completed The creation of an attitude reference system using a gyrosensor and accelerometer which estimates the inner frame angle has been created and tested although the accelerometer was damaged towards the end of the project and needs to be replaced Closed loop yaw rate control The actual yaw rate shall be measured and compared to the desired yaw rate and the error between the two shall be minimized This shall require speed sensors on the two drive wheels Two sensor options are encoders and DC motors Issues with accuracy at low speed may be encountered with the encoders Due to the problems arising when implementing motor speed feedback system using encoders and time limitations the rotational speed of the motors was not able to the fed to the MCU Thus the yaw rate was not able to the measured and closed loop yaw rate control was not achieved An encoder system to measure motor speed has been implemented on the diwheel using photo detectors to sense infrared pulses broken by holes in the brake disc coupled to the rotor of the drive wheel However this system was not successfully integrated with the control programming in time for
142. gnal is sent to the Dragonboard indicating that power has been removed from the motors The emergency shut off circuit is powered from 24V while the relays are fed 12V and require around 0 5A in order to latch In order to return power to the motor controller once the emergency stop button has been pressed the stop button must be deactivated followed by the activation of the start button The Power Control Board also accomodates the battery monitoring circuit which provides the Dragonboard with a signal proprtional to the voltage level of the batteries This is used in an attempt to maintain maximum responsiveness of the motors as the battery voltage drops The battery monitoring cicuit consists of two resistors in series forming a voltage divider outputting an analog signal bewteen OV and 5V to be read through an ADC on the Dragonboard The resistor values were chosen such that OV into the Dragonboard represented OV across the batteries and 5V into the Dragonboard represented 26V across the batteries as each of the 2 batteries may be charged up to 13V The resistors chosen for the battery monitoring system were Rag 12k0 In the actual implementation only two thirds of the bus voltage is measured by the voltage divider The assumption is that the estimated voltage level of fraction of the battery voltage is proportional to the entire voltage The battery voltage may be viewed on the Dragonboard LCD through use of the menu system in contrast t
143. h data Call SD_writeSingleBlock x where x is the number of the 512 block of data to be written to e For reading call SD_readSingleBlock x where x is the number of the 512 block to be read The data read is stored in the global array buffer The code used to interface with the SD card via the SPI protocol was modified from code available at www dharmanitech com 117 8 Software Design 118 9 Experimental Results and System Integration The following chapter displays and discusses the experimental results derived from testing the control systems implemented on the physical diwheel This chapter also contains descriptions of solutions to problems encountered in the integration of EDWARD s electrical software and hardware systems 9 1 System parameter identification To provide an accurate mathematical model of the diwheel several parameters incorporated in the dynamics of the diwheel needed to be determined Many of these parameters were determined from the 3D CAD model of the diwheel after care was taken to ensure material properties of components were accurately applied in the model These parameters included the mass and centre of gravity of each body of interest along with the moments of inertia of each body about the rotational axes slosh and yaw axes Parameters that were unable to be found from the CAD model were the damping terms associated with the rolling friction between the inner frame and the outer wheels
144. he brake is t The shear force on each of the bolts in one plane of shearing is V The bolt steel is assumed to be of the properties of 4 6 6 bolting category of the minimum tensile strength fuf 400M Pa The radius of each hub bolt from the center of rotation is r 20mm Figure B 2 The three hub bolts on the rim of the mini moto drive wheels From Australian Standard of Steel Structures AS 4100 the design sheer capacity is given as V 0 62 fufkr NnAc NrAo Where e For a lap length of less than 300mm k 1 e There are n 0 shear planes through the thread 157 B Appendix e There are n 1 shear planes through the shank of area A 12 566mm e The design capacity factor is y 0 8 The design capacity of each bolt is then eV 0 8 x 0 62 x 400 x 1 x 1 x 12 566 2 49 NV When applied to the three hub bolts there is a design torque capacity of t 3Vr 3 x 2490 x 0 02 gt t x 150Nm When compared to the motor torque and maximum braking torque of approximately 14Nm and 10Nm respectively the hub bolts seem adequate B 3 Mechanical Design Dynamic Analysis To initially estimate the distance of the center of gravity from the center of rotation of the diwheel a simplistic energy approach can be undertaken by assuming that all of the diwheel s kinetic energy traveling at 8km hr is converted into potential energy in inverting the passenger by locking the brakes The following terms are used in this calculation
145. heel This design is used in THE TRINITY Figure 2 5 A possible problem with a friction drive is that the driving torque is limited by the possibility of slipping between the outer wheel and the drive wheel The source of propulsion is commonly a petroleum engine featured in the designs of Vereycken Figure 2 3 Kerry 2 Background Mclean Figure 2 4 and Bressen Figure 2 7 The diwheel by Edouard Vereycken Figure 2 3 is differentially driven by a petrol engine The differential allows the use of a single engine by distributing the propulsion to each wheel while enabling the independent rotation of each wheel through a braking system Other forms of power include pedal as in the monocycle by Ben Wilson Figure 2 11 and electric motors as featured in the Moovie Figure 2 2 and the Trinity Figure 2 5 2 1 6 Steering Steering of the diwheel requires the separate and controlled drive of each wheel and has been achieved through many methods The steering of Edouard Vereycken s diwheel is controlled by a steering wheel rotated by the passenger much the same as in an automobile Vereycken s design featured brake shoes housed within each drive wheel which when the steering wheel is turned are activated accordingly to slow the drive wheels This then governs the distribution of propulsion though the means of a mechanical differential Vereycken 1947 The Peugeot Moovie Figure 2 2 uses variable control of the electric drives on each of
146. heel and therefore form the foundation of the load support As such care must go into the design of this system to ensure minimum rolling friction maximum drive wheel traction and structural integrity 58 6 3 Outer and Inner Wheel Design 6 3 1 Design Criteria and Aim The principle design function of the diwheel system is to carry one person in the forward and reverse direction and stands as a concept solution for a future transportation vehicle The design of a diwheel is concerned with three main aspects 1 Safety Safety is the one of the primary elements of the design of a consumer vehicle The structure must be robust and durable In order to protect vehicle occupants proper harnessing should be included and the vehicle structure should be designed to withstand and absorb all or part of energy in the event of an impact The design aim for safety is to provide for the securing of a harness to the inner frame and also proper constraint for the user if the diwheel is allowed to become inverted 2 Function and performance Basic function and performance include the diwheel s top speed its stability and the comfort afforded to the user The initial requirements of these performance measures include a minimum operating speed of 10 km hr the ability to perform steady turns at different radii and the capability to reduce the driver sensation of road disturbances 3 Cost Cost in a design project is one of the principle drivers of the proje
147. i 2005 Y G Martynenko and A M Formal skii Theory of the control of a monocycle Applied Mathematics and Mechanics 69 4 516 528 2005 Martynenko 2007 Y G Martynenko Motion control of mobile wheeled robots Journal of Mathematical Sciences 147 2 6569 6606 2007 Readman and Wellstead accessed Aug 2009 M Read man and P Wellstead Ball and hoop 2 Con trol and analysis Technical report Controlsystems principles co uk accessed Aug 2009 Ruan and Cai 2009 X Ruan and J Cai Fuzzy back stepping controllers for two wheeled self balancing robot In International Asia Conference on Informat ics in Control Automation and Robotics 2009 Self accessed Aug 2009 D Self Di cycles and diwheels Online www dself dsl pipex com MUSEUM TRANSPORT accessed Aug 2009 Wang and Fang 2004 Chen Y Wang Z and N Fang Minimum time swing up of a rotary inverted pendu lum by iterative impulsive control In American Con trol Conference pages 1335 1340 IEEE 2004 Wellstead accessed Aug 2009 P Wellstead Ball and hoop 1 Basics Technical report Control systems principles co uk accessed Aug 2009 Yamakawa 1989 T Yamakawa Stabilization of an in verted pendulum by a high speed fuzzy logic controller hardware system Fuzzy Sets and Systems 32 2 161 180 1989 Yoshida 1999 K Yoshida Swing up control of an in verted pendulum by energy based methods In Amer ican Control Conference
148. igure 7 15 A comparison of the open loop step response between the motor voltage and the slosh angle and the closed loop controlled step response is shown in Figure 7 16 It can be seen that with the slosh controller implemented the oscillations of both the slosh angle and slosh angular rate have been significantly reduced and the inner frame has reached its steady state position following a trajectory that is much more comfortable for the driver 97 7 Control System Design Root Locus for Slosh Angular Rate Imag Axis o 35 8 1 5 i 0 5 0 Real Axis Figure 7 14 Root locus for voltage to slosh angular rate transfer function showing closed loop pole locations indicated by pink squares Step Response 0 03 1 1 Open Loop Slosh Anglular Rate tc Closed Loop Controlled Slosh Angular Rate 0 02 0 01 Amplitude 0 01 0 03 0 04 gt 8 Time sec Figure 7 15 Closed loop step response of open loop slosh angular rate for voltage input and closed loop with slosh controller gain K 31 98 7 6 Inversion Control Step Response 0 005 0 01 0 015 Amplitude 0 02 0 025F Open Loop Slosh Angle ang Closed Loop Controlled Slosh Angle 0085 5 10 15 20 25 Time sec Figure 7 16 Closed loop step response of open loop slosh angle for voltage input and closed loop with slosh controller gain K 31 7 6 Inversion Control The pu
149. in he aol e Boe axe GER ae A Battery e ok wun Se a oe ae oe ee eon Be ee BR ee ey E 6 1 6 Joystick Mounting DEl ois Eo Rae BS ee Pdo nL OS 6 2 Finite Element Analysis qh sage ie ae a 6 2 1 Idler Wheel Contact Problem o o 6 2 2 Idler FEA Results and Conclusions 6 2 3 Space Frame Analysis e pos dos wi a ae Sa Be 6 2 4 Space Frame FEA Results and Conclusions 6 3 Outer and Inner Wheel Design aora dadas Pee AS 6 3 1 Design Criteria and Aimer oy ee ley due ea pe hs a pe ie Qe eS 6 3 2 Concept Solutions oa denon e A a Ga ee 6 3 3 Final Design of the Inner Wheel 2 2 6 se bao Yd eS ee eA Re 7 Control System Design 7 1 Analytical Derivation of the System Dynamics TAT CMechameal Dynamics esis goanen a a oa e e HRA A 7 1 2 Euler lagrange Derivation rada et SS ES a ES ra Electrical Dynamics ep ae ee ie Pe e ind pb ls sa 7 1 4 linearised Dynamics 64 45 44 40 a A A oe wet ees 7 1 5 Summary of Assumptions ra Seve Rae ak Se Se Ae A RY 7 1 6 Model Parameters ts a dd 2 A POS beg Ad td 7 2 Simmechanics Derivation of the System Dynamics 221 Main Block Diagrami ie io yo si wos A a a T22 AMET PTAS e e e Sake a Shae a le dd 37 38 40 41 41 42 44 46 49 50 50 52 52 53 54 54 54 54 56 57 57 58 59 59 67 Contents dz Drive System es E e E NA A AA de ke Mek 83 TZ A Outer Wheels eea Ne oe a o i se de o a a WON s 85 Lio
150. ing accelerometer Bot tom green Implementation on the Dragonboard Bottom blue Kalman Filter in MATLAB series method and cannot be altered by the user For this reason a two argument arctangent atan2 function was created for use on the dragonboard The implementation adapts the method presented in Lyons 2005 pp 547 549 which divides the unit circle into octants and uses an approximation of atan2 for 0 45 with a maximum error of 0 26 Two such atan2 functions were implemented in code a fixed point version and floating point version The functions can be found in angle h 8 6 Implemention of the Control Systems Although time constraints have restricted the implementation of all the controllers described in Chapter 7 a basic proportional feedback controller was implemented to provide slosh control The idea was to simply feed back a control gain proportional to the slosh angular rate using negative feedback By changing the sign of the gain and increasing its magnitude an effective swing up controller was produced which relied on making the closed loop poles of the system unstable The results of both of these controller are discussed in Chapter 8 6 1 Slosh Control The form of slosh control implemented to date is the simplest The pseudocode required for this control is 116 8 7 Additional Functions K sloshgain 31 0 8 13 K oshgain slosh 970 36 fostig 8 14 sloshcommand slosh x K 8 15 8 6
151. ing up for all but the most benign cases 5 Conclusion and Future Work In this paper the dynamics of the diwheel were de rived where it was seen that it exhibits behaviour seen in other nonlinear under actuated unstable mechanical plants such as the inverted pendulum and self balancing wheeled robots Fantoni and Lozano 2001 Conse quently approaches applied to systems with similar dy namics are also applicable to the diwheel which have been successfully demonstrated here Closed loop response 50 A 1 0 Po o e e e ee ee ee ee eee ee eee ee ee 50 9 degrees 3 Control Volts 3 E lt 100 150 A 2 i i i 005 5 10 15 20 Time Figure 10 Closed loop response of inversion controller to an initial pose of 04 170 Future work will involve extension of the two dimensional model to a fully coupled three dimensional model including yaw arising from differential motion of the wheels In parallel to this work will also continue on the EDWARD diwheel including the final assembly and commissioning conducting a thorough system iden tification testing the different safety systems and then finally implementing and benchmarking the various con trol strategies Acknowledgments The authors would like to acknowledge the support of the Bob Dyer and Phil Schmidt for their efforts in con structing the physical diwheel used as the basis of the model in this paper References Aboel Hassa
152. input making it cheap and relatively simple however it results in a control system that is not very robust and very sensitive to disturbance and modelling errors Aboel Hassan et al 2009 A more robust and effective form of slosh control uses sensors to measure system states and implements control loops to adaptively control the input signal sent to the plant actuators Al though this method requires more complex system equipment and a more comprehensive model of the dynamics of the system and its slosh production it provides much better suppression of slosh over a much larger range of system tasks and is much less sensitive to disturbances This control has been implemented in systems such as the ball and hoop system Wellstead and Readman and the two degree of freedom 2DoF slosh control rig Gandhi et al 2009 2 3 2 Relevance of Studies For the purpose of implementing slosh control on the diwheel the dynamics of the system during slosh need to be derived This then provides the ability to model the system and tune a control system to it Liquid slosh has been approximated as a suspended pendulum by Gandhi et al 2009 for their derivation of the dynamics of a two degree of freedom actuation slosh rig 12 2 4 Yaw Control and by Nichkawde et al 2004 for their derivation of the dynamics of a coupled slosh vehicle The Ball and Hoop system is also an approximation of slosh similar to a pendulum where the ball is the pendulum mass whi
153. ion due to the dynamics of the diwheel system Allowing for the accomplishment of the extension goal to remain upside down and considering that the diwheel may gerbil if large accelerations are allowed the seat and harness must afford proper constraint The seat and harness must work in combination to sufficiently restrain the occupant and must therefore be integrated to do so Furthermore safety is paramount in selecting a seat and harness and designing an integrated system due to the risk presented in the operation of the diwheel 25 4 Component Selection A regular automotive seat belt consists of a three point harness which outside of collision situations provides little support through a retraction system Racing harnesses consist of three four five or six point harnesses Figure 4 7 which are designed to provide constraint of the occupant during operational loadings Common safety problems involved with the use of a harness include submarining pressure injuries and spinal injuries States et al 1970 Submarining occurs when the harness is pulled up and the lap strap becomes positioned above the abdomen The occupant s pelvis is then allowed to move forward putting pressure on the soft abdomen The three and four point harness can lead to submarining because the abdomen may slide down beneath the belt Therefore a 5 or 6 point harness is required to constrain the passenger when sliding forward and to increase the restraint by holding the
154. ion required for this plant Dq Yes LEGISLATION OHS Legislation and or Australian Standard to be used for further reference OHS Regulations 1995 Divisions 1 2 OHS Responsibilities 1 3 Information Instruction Training Induction Supervision 2 10 Noise Part 3 Plant including 3 2 25 Plant with moving parts 3 2 15 Hazard Management a 3 2 17 Control of risk RISK ASSESSMENT TEAM Operator All project members HSO Manager Tao Zhu HSR Tao Zhu Other Other Risk assessment rating and action required Li No AS 4024 2006 Safety of machinery endorsed by Head of School Branch see also Other competency required AS 1788 Abrasive wheels apa DJ Yes AS 60204 1 electrical equipment for 11 10 09 No industrial machines Include in Safe Operating Procedure LF Tao Zhu Signature Name Page 1 of 10 Project No 757 Issue No 1 lt pss PROJECT RISK ASSESSMENT Page 2 of 10 Project No 757 Issue No 1 oe THE UNIVERSITY a F ADELAIDE IEE Ei PROJECT RISK ASSESSMENT LEGEND O El Elimination CL Su Substitution O En Engineering L is Isolation L Ad Administration Ll PPE Personal Protective Equipment Ll CAR Corrective Action Register Hazard Identified 1 Yes e g Hair Jewellery Clothing Cleaning aids cloth Gloves or Other No O O NA Likelihood Consequence Score Unlikely First Aid Low Comments
155. is 10 2 Cost Analysis Throughout the project cost was a major driving factor in the selection of components and manufacturing techniques Each group member is entitled to 200AUD from the University of Adelaide for use in the final year project With five students in this group the project budget without external sponsorship was 1000AUD After external sponsorship was sought only few parts were donated to the project or sold at a reduced cost and no monetary sponsorship was obtained despite efforts A summary of the costs incurred throughout this project is given in Table 10 1 which in cludes the component and manufacturing costs Parts which were obtained at no cost due to sponsorship or the recycling of parts from previous final year engineering projects are shown in italics but are not included in the total cost In the construction of the physical system a large amount of labour was required from both the mechanical and electrical workshop at the University of Adelaide A summary of the total workshop hours attributed to this project is give in Table 10 2 and summary of the working hours of each group member is provided in Table 10 3 133 10 Final Design Analysis 134 Table 10 1 Cost of parts for the EDWARD project Category Part Outer Wheel Inner Wheel Transmission Space Frame Brakes Miscellaneous Processor Power source Sensors Speed yaw control Other Electrical Tyre materials and bonding Rolled
156. is zero This means that the second state y is unobservable and cannot be estimated purely based on the gyrosensor and accelerometer measurements i e and 9 respectively However because the balancing controller does not require y this is not a big issue and an observer could be designed for the system which determines y based on the two sensor measurements Such an observer has not been designed at this stage but would be a valuable future addition to the system especially if measuring the drive wheel motor speed was problematic 106 8 Software Design This chapter describes the software as implemented on the DRAGON12 PLUS processor includ ing the design methodology inputs and outputs required by the system and techniques used to process sensor data Finally a number of useful non project specific additions are described which improve the system s flexibility While techniques of software implementation using a REAL TIME TARGET or use of a DSPACE controller have been used successfully in past projects for implementation of control on a mi croprocessor it was decided early in the project that all software development would be done in C or assembly programming the microprocessor directly The reason for this was chiefly in the added control over the compilation process The REAL TIME TARGET a MATLAB toolbox written by Dr Frank Wornle formerly of the University of Adelaide which converts a SIMULINK model directly into runnable C
157. ition 0 7 0 y y 0 gives the linear state equations AUDA 0 0 ar Ji J2 0 E 1 y 0 0 0 ar Side ge Sedo Jag 0 J ar b 2 J ar b 2 arb Ag QR 0 Ja y ar b z Ja ar by2 Job 1 and 0 1 0 gt 7 34 ar Jy Jo J ar Ja ar The poles of this plant are at s 0 0 25 2 27 3 07 Note that for the case of a voltage input u V then the damping term arising from the differential velocity of the frame and 80 7 2 Simmechanics Derivation of the System Dynamics wheel increases from b12 b12 bm resulting in open loop poles at s 0 0 34 2 12 3 40 and the state input matrix B needs to be multiplied by Nisin as was done for the downward R linearisation case 7 1 5 Summary of Assumptions In deriving the dynamics of the diwheel several assumption have been made to simplify the model A summary of these assumptions is given here e The motion is restricted to the ry plane e Friction is limited to viscous friction and the Coulomb friction arising from the idler rollers is neglected e The suspension arms are fixed keeping the centre of gravity of the inner frame a fixed distance from the centre of the wheels e The inductance of the motor is negligible and therefore the current is an algebraic function of voltage and motor speed e The rotational and translational inertia from the motors and drive wheels has been in cluded in the inner frame
158. ive system Causes partial loss of exclude Driving wheel mechnical power motor slip Outside rubber material wears quicker Driving wheel derail Completely loss of mechnical power Possible damage of motor Drive system mounting failure Over deformed by unknown external force To be operated by Disassemble and fix or replace authorized personel only the damaged part So THE UNIVERSITY a F ADELAIDE IEE EADE This form is to be completed in conjunction with the Plant Hazard Identification Form and in accordance with the Hazard Management Policy Procedure and Plant and Equipment Safety Management Policy Procedure IDI PROJECT RISK ASSESSMENT Overall risk rating existing controls highest score HIGH eg L M VH H STEP 1 ENTER INFORMATION ABOUT THE ITEM OF PLANT EQUIPMENT IT S LOCATION AND THE PEOPLE COMPLETING THE RISK ASSESSMENT Campus School Branch Building Room North Terrace School of Mechanical Engineering South Student amp Mechanical Date assessed 11 10 09 Engineering Engineering Workshops Plant Include name and EDWARD Electric Di Wheel Review date 15 10 09 model i SUr of Honours student project vehicle for driving and testing control structures AFFIX PHOTO HERE REGISTRATION LICENCES COMPETENCIES Refer Appendix B Registration required Yes IX No Licence Trade certificate required O Yes I No Supervisor assessment required Induct
159. ive This can be achieved by reading speed feedback from the motors and providing it to the appropriate closed loop controller e During strenuous testing the motors were noted to become very hot approaching 70 This was caused by their small form factor and inability to dissipate heat quickly enough Resolving this issue might involve implementing a passive or active cooling system or replacing them with motors that have improved thermal capabilities e An optical solution was selected to measure the speed of the motor however these sensors can be vulnerable to stray light and dust particles Incorporating a more robust sensor such as a magnetic solution or DC motor tachometer would eliminate these vulnerabilities e The implementation of inversion controllers as described in Section 7 6 would allow a better demonstration of EDWARD s full capabilities especially as a system for applied control e The implementation of an observer estimator on the Dragonboard used to provide an estimate of p without directly measuring it could be used in for small slosh angles reducing the need for the encoder sensors It could otherwise be used as a redundant speed estimate e The implementation of hydraulic brakes in order to eliminate differential braking has been considered as a viable option The implementation of a flexible sound system for reporting the diwheel s states to the driver as well as warning sounds and music e The impl
160. k 2006 who have designed a fuzzy multiswing swing up controller for a small inverted pendulum on a cart Their control input consists of a voltage applied to a permanent magnet DC motor which parallels the control available on EDWARD Because of the voltage input the acceleration of the pivot point is not constant for all motor speeds Instead the effective magnitude of acceleration was used to determine their control gains 99 7 Control System Design Balancing control as applied to inverted pendulum systems has been achieved using the linear quadratic regulator LQR Chang et al 2007 Lauwers et al 2006 Muskinja and Tovornik 2006 as well as fuzzy distributed pole assignment techniques which use fuzzy controllers to perform pole placement Tao et al 2008 This section will present the design of a fuzzy swing up controller and a LQR balancing con troller for use on EDWARD In further sections a discussion is given as to the ability to perform trajectory tracking when the inner frame is balanced upside down as well as issues with feedback of the entire state vector 7 6 1 Design of Fuzzy Swing up Controller This section describes the design of a fuzzy swing up controller The history and detail of the method behind the application of fuzzy control will not be treated in depth The authors recommend Jantzen 1998 as a good introduction to the design of fuzzy controllers Design of the fuzzy swing up controller was carried out usin
161. lman Filter Produces an optimal estimate of the states which minimises the process noise and sensor covariances Can track gyrosensor bias Only considers linear systems e Extended Kalman Filter Similar to the Kalman Filter but extended to non linear time varying state matrices Complicated The alternative to combining the two sensors together is to use an Inertial Measurement Unit IMU which contains gyrosensors and accelerometers and sensor processing within a convenient package However IMUs are very expensive devices and a more affordable solution is desirable 113 8 Software Design 8 5 1 Sensor Fusion using a Kalman Filter The gyrosensor and accelerometer have been fused using a simple Kalman Filter which estimates the inner frame angle and gyrosensor bias The two inputs to the Kalman Filter KF consist of the biased gyrosensor reading 09 and the angle estimate given by the accelerometer 04 The state equations used to describe the sensor states are given by Xiy1 Ax Bu 8 3 De 1 dt Oki de ea aieh e and Vivi Cx 8 5 t 1 ol n a k A E where dt is the time between gyrosensor readings wz_1 is the process noise and vz is the noise of the accelerometer measurements in radians As shown in Equation 8 4 the model of the states consists of the inner frame angle 6 and gyrosensor bias d The first step of the Kalman filter consists of the Prediction step which estimates th
162. lytical and SIMMECHANICS models restrict motion of the diwheel to one plane the two outer wheels are assumed by this model to be rotating at one speed Therefore the model combines the two wheels into a single wheel with twice the inertia of one wheel Figure 7 5 shows the basic block diagram At the left of the diagram the To Motor input which arises from a connection to the drive wheel joint connects to a body block that contains the mass and inertia of the two wheels To the right of this block are a Body Sensor and Rolling Constraint The body sensor is capable of measuring all coordinates and derivatives of the outer wheel However in Figure 7 5 the measured quantities are restricted to the wheel s linear velocity t and its angular velocity p The rolling constraint is used to tell SIMMECHANICS that the wheel s linear and angular velocities are algeraically related i e t Ry This stipulates that the outer wheel maintains pure rolling contact with the ground In reality the outer wheels will never slip before the drive wheel does because it is the traction between the drive wheel and the inner surface of the wheels that provides the reactive torque The feedback through the gain block models the rolling damping of the outer wheels 7 2 5 Results The reason for verifying the dynamics of the analytical solution is to ensure that any simulations performed with either set of dynamics are a realistic representation of the true physical syste
163. m The two DoF models described in this section and Section 13 are only representations of this diwheel s behaviour in the absence of any turning yaw motion about the y axis However for the purpose of the Slosh and Inversion control systems described in Sections 7 5 the models form a solid basis for simulation and controller testing 7 2 6 Visual Simulation Using VRML In order to assist the design process further the dynamic models derived in Section 13 and 7 2 were to be simulated in conjunction with a Virtual Reality simulation which formed one of the principle project goals The virtual reality simulation was achieved by utilizing MATLAB s VRML TOOLBOX The initial task involved exporting a simplified version of the final CAD assembly model to a WRL file which contains a purely text based representation of all 2D 3D 85 7 Control System Design primitives of the three dimensional model The second task required the grouping of individual primitives into assemblies of the diwheel and finally into transform groups which produces a part hierarchy so that the motion of a parent part has the correct effect on child parts The result is a file which contains groups associated with the following assemblies on the diwheel e Left and right wheel e Left and right drive wheel e Inner frame including the space frame seat idlers and spring dampers as well as other non critical parts This structure allows visual simulation of the diwhe
164. m Rix R 720 mm r 140 mm e 160 mm Damping bi 30 Nm s rad b 12 Nm s rad Motor Veas 36 V Rm 0 314 Ohms E 0 3 mHenry Km 65 mNm A Transmission N R 5 14 N 7 7 2 1 Main Block Diagram A view of the main SIMMECHANICS block diagram is shown in Figure 7 2 The main block diagram consists of three interconnected entities the inner frame assembly the combined two outer wheels and the drive assembly The connection between the inner frame and the outer wheels is assumed to be solely via the drive system consisting of the drive wheel and the motor and transmission 7 2 2 Inner Frame The block diagram of the inner frame assembly is shown in Figure 7 3 Beginning at the left of the diagram the block diagram consists of a Machine Environment block connected to a Ground block This is the root of the SIMMECHANICS model and there is only one connection to the Machine Environment block The physical constraints placed on the inner frame body consist of two degrees of freedom with respect to the Earth Ground centered reference frame one rotational revolute and one translational prismatic The two degrees of freedom describe the behaviour of the inner frame as being able to rotate freely within the outer wheels and translate parallel to the ground in a vertical plane The Body DOF block in SIMMECHANICS requires the axes about which the DoFs are applied Following the derivation in Section 13 the rotational DoF is applied about the z axis while
165. n and also maintain a constant contact force between the idlers and the wheel Two brushed DC motors each drive via sprockets and a chain a small motorcycle drive wheel which contacts the inner radius of the outer wheel Thus the vehicle can be driven forwards and backwards using a combined voltage applied to the motors and can be steered when the motors are driven differentially The vehicle was made drive by wire where the driver controls the vehicle via a joystick which 1 Introduction interfaces with a microprocessor A mechanical hand brake operates calipers on the drive wheels Three sensing systems were implemented a solid state gyrosensor for measuring the slosh rate of the inner frame a solid state DC coupled accelerometer for state estimation of pitch angle and incremental encoders on the two drive wheels The conceptual design of EDWARD was commenced in parallel with the design of an elec trical control system The open loop and closed loop control strategies which were to be implemented on a microcontroller unit MCU were designed following the derivation of the dynamics for the diwheel The dynamics which have not been presented elsewhere have been derived using the Lagrangian approach and SIMMECHANICS using a mechanical based ap proach Measured parameters of EDWARD were used in a virtual reality simulation for which the open loop response and various closed loop responses were simulated The concept of using feed back
166. n et al 2009 A Aboel Hassan M Arafa and A Nassef Design and optimization of input shapers for liquid slosh suppression Journal of Sound amp Vibration 320 1 15 2009 Astrom and Furuta 1996 K J Astrom and K Furuta Swinging up a pendulum by energy control In FAC 15th World Congress 1996 Cardini 2006 S B Cardini A history of the monocycle stability and control from inside the wheel EEE Control Systems Magazine 26 5 22 26 2006 Chang et al 2007 Y H Chang C W Chang C H Yang and C W Tao Swing up and balance control of planetary train type pendulum with fuzzy logic and energy compensation International Journal of Fuzzy Systems 9 2 87 94 2007 Dyer et al 2009 C Dyer K Fulton J Harvey E Schumann and C Zhu Electric Di Wheel with Automatic Rotation Damping Technical report The University of Adelaide 2009 Fantoni and Lozano 2001 I Fantoni and R Lozano Non linear Control for Underactuated Mechanical Sys tems Springer 2001 Grasser et al 2002 F Grasser A D Arrigo S Colombi and A C Rufer Joe A mobile in verted pendulum IEEE Transactions on Industrial Electronics 49 1 107 114 2002 Lauwers et al 2006 T B Lauwers G A Kantor and R L Hollis A dynamically stable single wheeled mo bile robot with inverse mouse ball drive In EEE Int Conference on Robotics and Automation pages 2884 2889 Orlando FL 2006 Martynenko and Formal ski
167. n forwards and backwards using a collective voltage in to the motors and can be yawed when the motors are differentially driven The vehicle is drive by wire and the driver controls the vehicle via a joystick There is also a mechanical foot brake which operates calipers on the drive wheels in case of electri cal failure There are three sensing systems on board viz a solid state gyroscope for measuring pitch rate a solid state DC coupled accelerometer for state estima tion of pitch angle and incremental encoders on the two drive wheels The scope of the project was to not only design and build the mechanical and electrical platform but to also implement several control strategies to modify the dy namics The first was a slosh controller with the pur pose of minimising the rocking motion that occurs as the vehicle is accelerated or decelerated when torquing the drive motors This was deemed necessary after viewing videos of monowheels and diwheels in operation If im plemented correctly it would also allow maximum decel eration of the vehicle if necessary which occurs when the centre of gravity of the inner frame is horizontally aligned Figure 2 Photograph of the EDWARD diwheel during construction to the centre of the outer wheels Another control mode was also considered with the aim to make the ride in the vehicle more exciting This involved a swing up con troller followed by an inversion controller The purpose of
168. nal velocity in x direction of the CoG of the inner frame assembly body 2 is Voz R Ge cos 0 7 5 and the corresponding velocity in y direction of the inner frame CoG is va Oe sin 0 7 6 The magnitude of the velocity of the inner frame CoG is thus 1 va R e cos 0 8 e sin 9 a 7 7 Kinetic Energies With the translational and rotational velocities the kinetic energies of each body and thus the entire system may be found The rotational energy of the outer wheels is Jg Eir q 7 8 75 7 Control System Design where J is the combined moment of inertia of both wheels about their centre The translational energy of the outer wheels is Rm Ey HE 7 9 Now the rotational energy of the inner frame is Jo 0 En E 7 10 where Ja is the moment of inertia of the inner frame about the centre of rotation of the diwheel Lastly the translational energy of the inner frame CoG is 1 Eo anaes ma Re e cos 6 6 e sin 6 gt 7 11 where N E is the ratio of outer wheel radius R to drive wheel radius r Thus the total kinetic energy of this system is Ex Erir Eu Bop Ex ma Re e cos 0 6 e sin 6 2 J p J 6 R p m4 2 2 2 7 12 Potential Energy As the outer wheels move only in the zz plane they maintain a constant potential energy Therefore the total potential energy assuming zero potential energy at 6
169. ncreasing interest as the implications of the problem have been found to be significant in systems such as satellites spacecraft and large ships Nichkawde et al 2004 2 3 1 Previous Work Although much work has been done on deriving the dynamics of liquid slosh for various con tainer shapes and fill levels Nichkawde et al 2004 Aboel Hassan et al 2009 Dodge 2000 Gandhi et al 2009 little work has been published on the active control of slosh Liquid slosh control has been attempted at ranging levels of complexity A basic form involves im plementing physical baffle plates within the liquid vessel to dampen the flow of the liquid and hence help to suppress the slosh Modaressi Tehrani et al 2007 Designs of this form are predominantly based on experimental results and utilise Computation Fluid Dynamics CFD software to analyse the slosh in a tank before and after the addition of various types of baffles Modaressi Tehrani et al 2007 A more complex form of slosh control involves the use of input shapers This method of control uses the flexible modes of a system to cancel out residual vibration by altering the reference signal sent to the actuator through the addition of self destructive impulses Aboel Hassan et al 2009 In reference to slosh control this method sends an altered speed control curve to the motor to prevent the excitation of the liquid slosh modes This does not require any form of sensory data to shape the control
170. nd Anathkrishnan N 2004 Stability analysis of a multibody system model for coupled slosh vehicle dynamics Journal of Sound amp Vibration 275 1069 1083 Ottoson A L and Lovsund P 1989 Protective level of safety harnesses combined with some racing car seats in frontal impacts a laboratory study Pergamon Journals Ltd Parks B 2005 Makers all kinds of people making amazing things in garages basements and backyards O Reilly California Riegel R 2003 Riegel s handbook of industrial chemistry Spring Press Ruan X and Cai J 2009 Fuzzy backstepping controllers for two wheeled self balancing robot In International Asia Conference on Informatics in Control Automation and Robotics Santana J Naredo J Sandoval F Grout T and Argueta O 2002 Simulation and construction of a speed control for a de series motor Mechatronics 12 1145 1156 Serious wheels 2005 Peugeot moovie concept Viewed 11 May 2009 Shigley J and Mischke C 2004 Mechanical Engineering Design 7th Edition McGraw Hill New York Southall D 2009 Dave southall s shed projects States J D G S and A F H 1970 The medical aspects of driver protection systems and devices developed through automobile racing Technical report SAE technical paper series Tao C W Taur J S Hsieh T W and Tsai C L 2008 Design of a fuzzy controller with fuzzy swing up and parallel distributed pole assignment s
171. nderneath the mass of the inner ring For large angles and large angular rates the voltage is zero to allow the inner ring to slow down for capturing by the inversion controller 3 3 Inversion control A linear full state feedback controller was used to keep the inner frame in the upright open loop unstable po sition Linear Quadratic Regulator A linear quadratic regulator LQR was used to stabilise the plant in its unstable position Section 2 3 This approach has been used successfully in the inverted pen dulum problem and its many variants Lauwers et al 2006 The cost function that was minimised is given by oS x 2 f 110 a 6 2 Rav dt 34 0 Table 2 Fuzzy rules Rule 1 If 8 PS amp 0 PS then Vm NB Rule 2 If 0 NS amp 9 NS then Vm PB Rule 3 If 8 PS amp 9 NB then Vm PB Rule 4 If NS amp 6 PB then Vm NB Rule 5 If 9 PM amp 6 PB then Vm NB Rule 6 If 9 NM amp 9 NB then Vm PB Rule 7 If PM amp 0 PS then Vm NB Rule 8 If 0 NM amp 6 NS then Vm PB Rule 9 If 0 PM amp 0 NB then Vm NB Rule 10 If 9 NM amp 6 PB then V PB Rule 11 If 9 PB then Vm PS Rule 12 If 9 NB then Vm NS Rule 13 If 0 PB amp 6 PB then Vm Z Rule 14 If 9 NB amp 0 NB then Vm Z where q and qo a
172. ng a hair net a Finger rings and exposed loose jewellery eg bracelets and necklaces must not be wom Medical Alert bracelet must be taped if exposed Page 1 of 5 SOP No 1 Issue No 1 OE SAFE OPERATING PROCEDURE pd EDWARD DIWHEEL Controller layout Steering open PABMIOT JAOJA PABMADE DAOJA System self diagnosis LJ Turn on the main switch LJ After 10 sec read feedback information on control board LCD Display All Test Passed gt All electronic and control states are normal Display Test failed 4 digit error code gt Do not operate the vehicle and report the error code to project staff A Do not press START button at this stage Page 2 of 5 SOP No 1 Issue No 1 Hs SAFE OPERATING PROCEDURE EDWARD DIWHEEL Operation Of Emergency Stop Button ODO OCOD OCODOODO O O Fasten 5 Point driver seat safety belt harness tightly after seated Ask project staff to switch power on at the main switch on top of the electric box and start the vehicle Press START button situated on the right side of the joystick controller Moving the vehicle for a small distance by pushing the joystick SLIGHTLY FORWARD The 5 point safety belt must be worn during this process Apply Emergency Stop Button and check that it is working correctly Check that the mechanical brake Emergency Stop micro switch is working Release the Emergency Stop b
173. nical brake Stop watch Dada 10 sec Braking time from full speed Electrical braking Stop watch Da 10 sec Open loop speed control Drive forwards Encoders Speed of the diwheel 10 km hr Drive backwards Encoders Speed of the diwheel 5 km hr The reduction of the sloshing Gyrosensor of the inner ring when 6 OO el accelerometer compared to the open loop zo response Stop watch Time elapsed while the Dynamically stabilized pitch gyrosensor diwheel s inner frame is 30 sec accelerometer inverted to 180 degrees Closed loop yaw control Encoders TE SES Or E 50 maximum yaw rate 19 3 Design Specification 20 4 Component Selection 4 1 Mechanical Brake Selection While electrical braking can be applied to the drive wheel via the motors in normal braking situations for added safety and for additional braking capability a mechanical brake can be used Brake systems often use frictional forces to apply a stopping torque to the drive system Disc and drum brakes are two types of brake systems common in modern vehicles Other types of brakes include caliper brakes used in bicycles which brake directly onto the rim of the wheel whereas disc and drum brakes operate within the drive system of the vehicle Hydraulic disc brakes were considered because of their ability to apply an even distribution in a hydraulic line to independant wheels which would eliminate the possibility of differential braking However hydraulics were dismissed as being to
174. njamin cazzolato adelaide edu au Abstract A diwheel is a novel vehicle consisting of two large outer wheels which completely encompass an inner frame The inner frame freely rotates inside the wheels using supporting idler wheels The outer wheels are driven from the inner frame and forward motion is achieved through a reaction torque generated by the eccentricity of the centre of gravity of the inner frame Dur ing operation diwheels experience slosh when the inner frame oscillates and tumbling when the inner frame completes a revolution In this paper the dynamics of a generic diwheel are derived Three control strategies are then proposed slosh control swing up control and inversion control Finally simulations are con ducted on a model of a diwheel currently under construction at the University of Adelaide 1 Introduction The diwheel is a device which consists of a two large outer wheels which completely encompass an inner frame The inner frame is free to rotate within the wheels and is typically supported by a common axle or idlers which roll on the wheels see Figure 1 Di wheels like their more popular cousins the monowheel have been around for almost one and a half centuries Self accessed Aug 2009 Cardini 2006 All of these platforms suffer from two common issues affecting driver comfort slosh and tumbling also known as gerbilling Sloshing is when the inner frame oscillates and it oc curs in all monowhe
175. o expensive and a cable drawn disc brake has been chosen The cable disc brake is intended to be used with a mini moto wheel hub and has been chosen in conjunction with the drive wheel for its simplicity and to reduce cost A photo of the purchased brake caliper is shown in Figure 4 1a A SolidEdge model of the brake disc rotor chosen is shown in Figure 4 1b As each drive wheel is rotating independently of the other independent brakes have been used on each drive wheel a Brake caliper b SolidEdge model of the brake disc rotor Figure 4 1 Mechanical brake chosen for use in EDWARD Brake Performance The performance of a disc brake system depends predominantly on the size of the brake pads and the radius and force at which the brake pads are applied to the brake rotor The measured specifications of the purchased brake caliper and rotor are shown following 21 4 Component Selection e Brake pad height 16mm e Proportion of the rotor the brake pad covers 18 degrees e Radius of the brake rotor 70mm The equation for the torque available from a disc brake with uniformly worn brake pads is shown below Shigley and Mischke 2004 p 831 1 TE 30 Or upari r r 4 1 The definition of the geometric terms of this equation are shown in Figure 4 2 The other terms are defined as e T Braking torque Nm e u Coefficient of friction of the brake pads e Pa Maximum allowable applied pressure on the disc pads Pa
176. o initial designs involving the voltage being dispayed on an external bar graph using a bar graph driver 5 4 Ignition Circuit During periods where the diwheel is not in operation it is necessary to implement a method of isolating the batteries and cutting off power to all electronics to prevent discharging of the batteries This is implented using an ignition circuit which consists of a key switch a relay and a circuit breaker The 400A key switch is located exterior to the control box and is the main point of isolation between the batteries and electronics The 100A circuit breaker is connected in series with the ignition and will trip removing power to the electronics in the event that current in excess of 100A is drawn from the batteries A relay inside the control box provides power to the Power Control Board once the ignition switch has been activated Al 5 Electrical Hardware Figure 5 5 Power control board which controls power to the motor controller and therefore motors Circuit breaker Figure 5 6 The key switch and circuit breaker which form part of the ignition circuit 5 5 Current Sensors The current sensors may be used to sense the motor current for feedback or monitoring How ever at this stage they have not been actively used in the design due to delays in the project Typical current sensor designs usually place a current sensing resistor either in series with the motor or in the lower legs of the H bridg
177. o not contain any thrust bearings This means the constraint plates on both sides will rub against it This is very unfavorable when the vehicle is operated at high speed since the both sides of center wheels will heat up due to 66 6 3 Outer and Inner Wheel Design friction causing increased wear Another major problem is that the actual load on the center caster wheels is only marginally below their designed capacity which may result in failure under small impact loads such as the load experienced when running over a small bump 6 3 3 Final Design of the Inner Wheel Design Overview Figure 6 25 Overview of the final design The final design possesses major improvements over previous suspension and bearing system arrangements and is shown in Figure 6 25 Radial movement of the suspension structure such as in Concept Design 3 where sleeves were used to constrain caster assemblies to move in outer wheel radial direction has been replaced by a H bridge bracket structure Ball bearing pivot joints are utilised for the suspension bridge and inner frame connection which provides excellent joint flexibility and load capacity Inner Wheel Design The inner frame has a triangular shaped profile due to a triangle s geometric stability The six sides are chosen to be 50mm RHS with a 2mm wall thickness according to RHS dimension selection and is given in Appendix The maximum load on each side of the wheel is estimated to be approximately 30
178. on Serious wheels 2005 Details of the system s design have not been made public However the concept design is described in a Peugeot press release Serious wheels 2005 and 2 Background can be seen in Figure 2 2 This highly stylised vehicle prototype is not a true diwheel as it is stabilised by two spheres which act as front wheels Figure 2 2 The Peugeot Moovie Serious wheels 2005 Patent archives contain many old monowheel and diwheel designs with details of their in tended mechanical realisation Detailed description of a diwheel concept and its working subsys tems is presented in the 1947 Belgian Patent Two wheeled vehicle supportive of each other Vereycken 1947 This diwheel was said to have been constructed and designed by Ernest Fraquelli Unknown 1935 and can be seen in Figure 2 3 Figure 2 3 A schematic from a Belgian patented diwheel Vereycken 1947 Dave Southall a senior lecturer in embedded systems at MANCHESTER METROPOLITAN UNIVERSITY in the United Kingdom has previously made a petrol powered monowheel and 2 1 Previous Designs is currently completing an electric diwheel Southall s diwheel is to the authors knowledge the only other electric diwheel in development Correspondance with Dave Southall has been useful for ascertaining detailed knowledge of some of the problems he has encountered with construction of his diwheel Other relevant constructions include the ROCKET ROASTER by Kerry Mclean and
179. on board and provides repeatable turning rates Following full operation of the diwheel the steering control was found to be intuitive allowing the driver to quickly learn how to control the diwheel Open loop speed control The forward and reverse speeds of the vehicle shall be controlled by analog data sampled from the steering system The vehicle s speed shall be measured by sensors and reported to the Mi crocontroller Unit for display to the user This user feedback shall include either LCD display 7 segment LED display or the use of a galvanometer The two former solutions are more likely as a speed estimate will be digitally stored within the MCU Open loop speed control was achieved with the desired speed command being measured from the joystick potentiometers The voltage across the motors is proportional to the speed input command allowing the diwheel to drive forwards and backwards Encoders were mounted on the drive wheel brake discs to provide a feedback of the current motor speed to the Dragonboard which can be used to estimate the diwheel speed The implementation of the speed feedback in the software was not completed due to project delays As such the open loop speed control goal has been achieved while the speed feedback to the user has not been completed Dynamic braking The motor controllers shall be capable of 4 quadrant operation This will be achieved by use of a H bridge driver The battery shall be capable of the curr
180. on view of the diwheel To make this simplification the left and right drive and outer wheels are each combined into a single entity with the assumption that both sets of drive and outer wheels rotate at equal speeds The schematic in Figure 7 1 shows a side on view of the diwheel representing the simplification into the single plane The coordinate frame of the system follows the right handed convention and is located at the diwheel s centre of rotation The z axis is defined as positive to the right in the forward direction of the diwheel and the y axis is positive down Rotations about the z axis are defined as positive clockwise The two masses m and ma represent the two bodies of interest of the diwheel where m is the combined mass of the left and right outer wheels and ma is the mass of the inner frame inlcuding a driver and the mass of the drive wheel assembly Both bodies have a moment of inertia J Jo about the centre of rotation of the diwheel To define the drive transmission the three radii are defined such that R R are the inner and outer radii of the outer wheels respectively and r is the radius of the drive wheel The eccentricity of the centre of mass of the inner frame assembly from the centre of rotation of the diwheel is given as e The model has three coordinates given by y the rotation of the outer wheels about the z axis 0 the rotation of the inner frame assembly about the z axis slosh angle and x the horizon
181. or the diwheel s speed remains unaffected This allows individual control of both the yaw rate and speed 7 4 1 Derivation of Yaw Dynamics In order to design a yaw controller an approximate mathematic model of the diwheel s yaw dynamics was developed The model has been simplified by assuming that inertia about the yaw axis is constant and that the inner frame does not slosh In fact if each drive wheel is driven at the same speed then no sloshing of the inner frame will occur The yaw rate 4 is achieved by rotating the left and right wheels at r and Yr respectively This meant that the yaw rate was controlled by the differential velocity supplied to both wheels by the motors This model was simplified further due to the symmetrical nature of the diwheel about the x y plane which allowed the two outside wheels to be grouped together Figure 7 10 shows a schematic used for the derivation The following symbols were used in the derivation of the yaw dynamics e A is the difference in angle between the left and right wheel yr yr e Va is the voltage applied to the right motor e Vz is the voltage applied to the left motor Va is the difference in voltage applied to each motor Vr Vz e 14 is the yaw heading angle 92 7 4 Yaw Control e Jw J Jg is the combined inertia of the wheels about their rotational axis e B is a viscous damping term proportional to the differential wheel velocity e F
182. orcycle cable drawn lever shown in Figure 4 3 The lever shown was subsequently modified to accommodate both cables from each brake and to distribute even tension between the two cables Signaling Operation of the Brake The application of the mechanical brakes must be registered by the processor in order to prevent driving whilst braking Braking the drive wheel while it is being driven will cause the control systems to sense a drop in speed of the wheels for which it must compensate by trying to increase the speed of the motors The solutions proposed to sense the operation of the brake lever include a normal micro limit switch and a Hall effect sensor The Hall effect sensor is a transducer which has an analogue output voltage which is altered in the presence of a magnetic field A Hall effect sensor may be used as a proximity switch due to its highly non linear output signal in the presence of a strong local magnetic field Such a magnetic field may be created by a small permanent magnet Because a limit switch would be invasive to the physical operation of the brake lever the simpler solution was to use a Hall effect sensor to sense the proximity of the lever arm when not in operation A Hall effect sensor shown in Figure 4 4 was set in epoxy resin on a small bracket that screws into the brake lever base and a small permanent magnet was set into a drilled recess within the brake lever arm see Figure 4 3 23 4 Component Selection Figure
183. pages 4045 4047 1999 A Publications Arising from this Thesis 154 B Appendix B 1 Hertzian Idler Wheel Contact The following solution for the contact pressures between two curves surface of different materials is an application of Hertzian contact theory The outer and idler wheel are considered as two elastic bodies in contact with two dimensions of curvature shown in Figure B 1 The terms are defined below e The material of the outer wheel body 1 and idler wheel body 2 are steel and nylon respectively e Young s moduli E 200GPa and E 2GPa and the reduced Young s modulus E e Poisson s ratio vy 0 3 and va 0 4 e Rj is the radius of curvature of body 1 in the x direction and if a curvature is concave the radius is negative e R is the reduced radius e The radii of the outer and idler wheel are are Ry 0 03m Riy 0 720m Roz 00 and Ro 0 04m e ko ki k3 k4 and ks are the contact coefficients e 0 90 is the angle between the direction x or y of the largest radius of each body The idler wheel design has two points of contact to support the overestimated 200kg of loading These surfaces are at an angle of 51 with the loading Therefore the loading is halved and is subject to a cos factor 2000 2 2000 N 2 The reduced Young s modulus is given by 1 1f 1 v2 1 Y 1 1 03 1 04 Ei 2 Ej Ez J 21 200e9 2e9 gt E 4 7109G Pa The reduced radius is given by 155 B
184. pt design of the joystick mounting is shown in Figure 6 9 The joystick mounting plate is connected to a pivoting tube which runs inside and is supported by two brass bushes The brass bushings clamp onto the pivoting tube and secure it in place using two bicycle seat quick lock clamps 6 2 Finite Element Analysis Finite Element Analysis FEA of the diwheel was used to analyse the contact between the tubular outer wheel and idler wheel and the space frame structure The following is a summary of the FEA performed on the diwheel 6 2 1 Idler Wheel Contact Problem The contact experienced between the tubular outer wheel and idler wheel is rolling contact between two elastic bodies However dynamic rolling contact was outside the scope of the investigation and so the contact was considered static The problem is then similar to a Hertzian 54 6 2 Finite Element Analysis Mounting plate Quick lock clamp Figure 6 9 Final design of the joystick mounting contact problem for which there are solutions between curved surfaced solid bodies in contact In this investigation the FEA model was developed from the conceptual design of the diwheel The material properties and loads were estimated and used in a model of the idler and outer wheel see Figure 6 10 in ANSYS the FEA software used The material of the outer wheel and idler wheel are Stainless Steel and Nylon 6 6 respectively The important properties of a material in the FEA and cal
185. put was available Consequently the vehicle was unable to counter the slip between wheels and driving surface As such there were variations between the desired yaw rate and the actual yaw rate and for this reason a closed loop system was implemented The closed loop system received feedback through a visual dead reckoning method using two optical mice mounted on the bottom of the robot which provided x and y displacement and the yaw angle as feedback This method when combined with a PID controller produced an improvement of two orders of magnitude in the yaw angle step response when compared to the open loop However this type of feedback is not viable for the diwheel as the outer wheels and inner frame are free to rotate relative to each other Baker et al 2005 constructed a co axial two wheeled Segway vehicle that used encoders for yaw control feedback Measuring the angular velocity of each motor allowed a yaw rate to be estimated This method of feedback proved successful in keeping the two wheeled segway travelling in a straight line 14 3 Design Specification EDWARD is a complex mechanical and electrical system containing many components that closely interact with each other Figure 3 1 shows a basic System Hierarchy Diagram of the diwheel system and its components Proper integration of all of the system components is important to achieve the project goals In order to achieve these goals each subsystem has an associated functional r
186. r measurement more than the gyro and noise susceptibility was increased although the response of the filter was good Making the magnitude of the diagonal terms of Q smaller but keeping their ratio the same allowed the filter to smooth noise in the accelerometer estimate of the angle using the atan function though small steady state errors between the filter estimate of 0 and the accelerometer estimate were noted especially when the angle was changed rapidly Therefore smaller terms in Q tend to make the Kalman Filter less capable of tracking fast changes in 0 The R matrix which is simply the measurement noise of the accelerometer was set at 0 9rad The main issue with the filter implementation is that the Q and R matrices are hard to choose realistically especially from a quantitative perspective In future a thorough analysis of these parameters should be taken to ensure the filter performance is a robust as possible A MATLAB script was written to receive data from the processor which consisted of the arctangent of the accelerometer axes readings and the Kalman Filter estimate of the angle of the sensor board both of which were calculated in floating point format The MATLAB script was also paired with an artificial horizon which provided visual feedback of the sensor board angle Figure 8 5 shows the performance of two Kalman Filters one implemented on the processor green and the other implemented in MATLAB blue These plots were produ
187. re continually being pro duced and improved upon as outlined in Chapter 6 The final CAD design was chosen on the basis of cost ease of manufacture ergonomics and aesthetics The CAD models presented indicate that this goal has been accomplished Build a diwheel to be driven by two DC motors including a frame to support one person and system equipment This goal will be verified by a final demonstration of the system concept according to the project outcomes The structural integrity of the frame shall be verified by performing FEA incorporat ing nominal loadings Proceeding from the completion of the final designs of the diwheel construction of the physical system was commenced This goal has been achieved as illustrated through the construction of a fully operational the diwheel presented in Chapter 6 The results of the FEA of the space frame is also provided in this section verifying the structural integrity of the frame Implement a mechanical braking system The diwheel shall be capable of stopping safely at top speed without assistance from the motors The mechanical braking system shall consist of two disc drum or caliper brakes each capable of stopping a wheel The mechanical brake shall be capable of stopping the vehicle during a loss of electrical power A mechanical braking system involving two caliper brakes on on each drive wheel has been implemented on the diwheel providing mechanical braking in addition to dynamic electri
188. re the state penalties on the inner frame angle 0 and the difference in the angular velocities be tween the frame and wheel 6 y respectively The pur pose of the latter term is to restrict the speed of the drive motor which is rated to 2100RPM The term R is the penalty on the drive voltage Using Bryon s rule the state penalty matrix for the state vector given in Section 2 3 was set to qa 0 0 0 36 0 0 0 g 9 o 0 _ o o o 00 q 0 0 0 0 q2 gq 0 0 3 35 and the effort penalty was R za 36 Solving the continuous algebraic Riccati equation re turns the optimal control gains k 193 0 44 11 37 Pole placement expensive control In parallel with the LQR controller described above a state feedback regulator was developed by using the common method of pole placement The placement involved reflecting all unstable poles about the imagi nary axis so that they were stable Note that this is equivalent to expensive LQR control where the penalty on the control effort is infinite The poles of the sys tem were moved from those given in Section 2 3 viz s 0 0 34 2 12 3 40 to s 0 0 34 2 12 3 40 resulting in a control vectors k 91 0 30 13 38 4 Numerical Simulations Numerical simulations were conducted on the differen tial equations derived in Section 2 within SIMULINK To validate the differential equations and the implementa tion within SIMULINK a parallel model
189. removing defects or detecting failure modes CRIT Criticality S x O RPN Risk Priority Number SxO xD Failure Mode Effect of Failure S R of o Current Process Controls D CRIT Iner Rotation Frame Joining Frame Casters Excessive deformation of outer ring structure can Outer wheel ring cause severe vibration of deform inner frame Inner frame excessively Jape jams if the ring has poor circularity Out Side rubber Decrease on max friction face demage E O Slip while turning or braking Distortion of all frame Too much i components deflection on p SHS Casters lose contact with outer wheel Bolt connection Structure failure and possible Severe vibration while Bearing failure running Caster wheels fall off the wheel shaft Inner stucture losses support Caster arm weld and could fall out if the failure broken caster is at critical location Caster Caster loss contact with 10 suspension fail wheel Design load is bigger than Excessive load on the inner actual applied load frame or poor travelling 1 Include safety factor surface condition Aviod operating on rough surface Poor road condition Oxidition a eae on bumpy Rubber metal bonding Store in cool place failure Excessive load on the inner Introduce design load and frame or poor travelling E safety factor surface condition Butt weld failure peruca failure and possible 10 Eee loa
190. rom an industry I beam through rolling process The large moment of inertia of an I beam makes this outer wheel design extremely robust and durable The caster assembly design enables casters to be locked in both directions inwards and outwards in outer wheel radial direction so that derailment where the inner frame loses connection with outer wheel is impossible However the fatal problem for this design is faced during manufacturing Through advice from the project Supervisor and rolling companies it was quickly discovered that rolling an I beam is complicated and cannot be easily achieved without structural distortion This is because the inner flange of the beam is subject to buckling when rolled In addition the rollers required for the caster assemblies were unable to be sourced Since this design contains major risks in manufacturing phase no further design improvements were made to this concept Design Overview and Description of Concept Design 2 Concept design 2 as shown in Figure 6 16 uses a rectangular inner frame One idler roller red is located at each corner to support to the frame and allow the required rotational movement while constraining the frame in the outer wheel axial direction The frame beam members are formed by Rectangular Hollow Sections RHS and connected by trapezoidal gusset plates with bolts Instead of bolt connection weld joints were also considered as a possible solution The profile of idler roller and roll
191. rom damper springs The bending moment on suspension bridge due to vehicle turning contributes mainly to thrust load on the bearings Some of the dynamic loads are not possible to predict without testing on the real vehicle structure so a large safety factor of 5 has been included in the bearing selection process From the load analysis the load condition on bearings has been classified as a combined load According to SKF bearing selection criteria Appendix deep grove ball bearings were selected All bearings for the swing arm pin joints and suspension pin joints are chosen to be 37mm outer diameter 12mm inner diameter and 12mm boss width Selection is based on the bearing static and dynamic load capacity refer to SKF deep grove ball bearing catalog Idler Roller Design The idler rollers are a critical component in the system as they support inner frame and allow it to rotate within the outer wheels Therefore the rollers have to be rigid and durable Nylon has excellent abrasion resistance and is mechanically strong Riegel 2003 which makes it a perfect material for the idlers Roller profile is calculated based on force distribution The roller surface profile was divided into 3 zones A B and C as shown in Figure 6 29 Zone A is the contact surface between the idler roller and outer wheel It has a wall angle of 39 69 6 Mechanical Design Figure 6 28 Final design Bearing selection degrees The wall angle is calculated a
192. rovide finer control to the inner frame angle A single voltage output membership function was used consisting of five singleton sets corresponding to 36 18 0 18 36 Standard names used in the design of FUZZY CONTROLLERS were given to each of the input and output membership functions such as POSITVE SMALL PS NEGATIVE BIG NB POSITIVE MEDIUM PM and the sign of both inputs followed from the reference frames assigned in Section 13 The angle input 0 was divided into six membership functions 100 7 6 Inversion Control Membership functions for Input Membership functions for de dt Input NM NS PS PM NS PS Degree of membership Degree of membership o o a o S o N Figure 7 17 Fuzzy controller Membership functions e Negative Big NB e Negative Medium NM e Negative Small NS Positive Small PS e Positive Medium PM e Positive Big PM and consists principally of three types of angle small medium and big Small angles correspond to angles close to the stable equilibrium of the inner frame the state of the inner frame at rest Medium angles are close to the 90 positions and big angles are close to the 180 inverted position The limits used in Muskinja and Tovornik 2006 were taken as a starting point Rules The Rules used in the fuzzy swing up controller are given in Table 7 2 All rules follow the format of If 9 PS AND 6 PS then Vm NB 101
193. rpose of inversion control is to swing the inner frame upside down and keep it there Typical approaches to this task have divided the problem into separate swing up and balancing controllers Typically the swing up controller is used to pump energy into the system so that the energy of the inner frame or pendulum is increased to that of the upper equilibrium Astrom 2000 describe such an energy pumping approach achieved by differentiation of the Lagrangian of a single pendulum system A non linear control law of u ng sign E Ep cos 0 was used which increased the pendulum s energy towards Ey at the inverted position While energy based control laws appear frequently in the literature they are not always amenable to the problem of swinging up a mechanical system because differentiation of the Lagrangian often produces a complicated control law Swing up can also be achieved using such strategies as fuzzy control Chang et al 2007 describe fuzzy controllers for both the swing up and balancing control of a planetary train type inverted pendulum Their fuzzy swing up controller consists of pure bang bang control with only two rules and two membership functions However such a fuzzy controller cannot be used on the diwheel because the magnitude of torque applied to the inner frame is decreased due to the gravity spring as the angle of the inner frame increases A more complicated fuzzy swing up controller is presented by Muskinja and Tovorni
194. rts being outsourced The final construction varied slightly from the initial concept design where certain details were incomplete or could be manufactured more easily another way The drawings created in order to construct the diwheel are found in Appendix D Figure 6 1 A photograph of the constructed diwheel The features of the final conceptual design of the diwheel are labeled in Figure 6 2 The concept design includes a tubular outer wheel driven by traction from the drive wheel The electric motors turn the drive wheel through a chain transmission The outer wheel is guided and supported by nylon idler wheels and a suspension system The space frame includes mounting locations for a racing seat and harness as well as a battery and electronics box The joystick is attached to a retractable mounting plate and the brake lever is secured on the space frame The mechanical design of the diwheel was separated into two sections The first section consists of the mechanical structure and components that exist between the left and right wheels named the space frame The second section includes the outer and inner wheel design including the drive and suspension system The detailed design of these two sections will be discussed in this chapter 49 6 Mechanical Design Figure 6 2 Labeled render of the final diwheel concept design 6 1 Space Frame Design The space frame of the diwheel structure is a rigid network providing support for the p
195. s and the line connecting the axis and the combined Center of Gravity acts as a virtual rigid link provided the Center of Gravity remains static relative to the inner frame In order that the upper unstable equilibrium point is controlled several tasks must be addressed e The Center of Gravity must be moved from the lower to the upper equilibrium e The Center of Gravity must be balanced at the upper position e The diwheel must maintain balance against disturbances including those induced by trajectory following Because the use of the inverted pendulum in control systems is well established many of these issues have been addressed in the literature A discussion of the control to be used in the proposed inversion controller is given in Section 7 6 11 2 Background 2 3 Slosh Control The implementation of control in the project aims to diminish the rocking of the driver within diwheel inner frame which occurs during harsh acceleration and deceleration This rocking motion is similar in dynamics to the sloshing of liquid that occurs in containers of fluid un dergoing lateral acceleration and deceleration Liquid slosh is evident in many systems such as vehicles containing tanks of fuel submarines and rockets When the vehicle accelerates or brakes the fuel undergoes oscillatory motion until the acceleration ceases and the liquid slosh decays to equilibrium The effects of liquid slosh on a system s dynamics have been investigated with i
196. s of the H bridge However frequent switching of the transistors due to PWM creates noise transients which are fed through to the common differential amp Dyer et al 2009 present a discussion of two amplifier ICs LTC6103 and LTC6014 produced by Linear Technology which provide a bidirectional load monitoring circuit with differential output or single ended output respectively Two representative circuits using these amplifiers are shown in Figure 5 7 However due to the high currents associated with high motor torques heat dissipation from the sense resistor can easily exceed hundreds of watts This requires the use of large expensive sense resistors which significantly degrade system integration The solution is to measure the current indirectly using a hall effect sensor These sensors make use of the principle of electromagnetic interaction by passing the load current carrying wire through the sensor The magnetic field produced by the current carrying wire induces a smaller proportional current in windings of the hall effect sensor which can then be converted to a voltage by passing it through a resistor Figure 5 8 shows the current sensor board of which there are two mounted on the diwheel Each sensor is capable of measuring currents ranging from 150A to 150A and requires a supply voltage between 12V and 15V The current sensor board uses a INA126 differential amplifier and ICL7660S charge pump IC for providing the required negative voltage
197. s of the candidate types are analysed by Dyer et al 2009 which guided the decision on the final battery selection A summary of the battery types considered is shown in Table 4 1 which has been compiled using battery specifications supplied by an online battery retailer and a battery selection guide Jaycar 2001 The main considerations for the battery selection were cost and capacity SLA Gel batteries provided the cheapest solution which met the needs of the project and were selected for use in the diwheel These are shown in Figure 4 12 Each of the four batteries purchased were 12V Gel SLA batteries rated at 26Ah and weighing 8 2kg each Table 4 1 Summary of Battery Properties Battery Mounting peta ae aes eines High Simple Very large Cheap Very heavy High High Simple Very large 360 Cheap SLA Gel Very heavy Low Medium Complex Large Medium 360 Expensive NiMH Weight Low Low Complex Large Medium 360 Expensive NiCad Weight Low High Complex Small Light 360 Expensive LiPo Low Medium Complex Small Light 360 Expensive Li ion Low Very Complex Very Small 360 Very High Light expensive LiFePO4 4 9 Processor In order to control the diwheel a controller in the form of a microprocessor is required The processor chosen for the design was the M9512DG256 manufactured by Freescale Semiconduc tor This microprocessor is utilised in the DRAGON12 PLUS known as the Dragonboard a development board produced by WyTEc The
198. s only rated at a maximum voltage of 43 Volts where the intended bus voltage was 48 Volts By removing one battery from the series connection the effective bus voltage was reduced to 36 Volts Simulations of the diwheel at this voltage were conducted and 125 9 Experimental Results and System Integration it was found the diwheel would still be able to achieve inversion It was decided to not remove the unused battery as it provided a useful distribution of weight Creating an effective encoder on the brake disc from two optical sensors was problematic The disc brake s misalignment caused fouling between it and the optical sensors This was overcome by removing excess casing from between the transmitter and receiver of the optical sensors Reading the brake disc did not provide equal on and off periods which would make the second sensor redundant if the sensor was not correctly spaced This meant directional feedback was an issue The encoders were sampled in succession with a constant elapsed time method This method provided accurate results during individual testing However it uses multiple interrupts and as such inhibits the main functions of the Dragonboard including speed and yaw control Consequently due to this effect and time delays they were not implemented into the main program Figure 9 7 Motor controller board destroyed 9 4 Hardware Issues The initial design of the diwheel proposed that the idler wheels be exposed with no ext
199. s segmented for the mounting of functional elements This design is also used in diwheels see Figure 2 5 for the function of providing a circular frame structure If the inner frame is restricted from rolling by mechanical means then the function of the inner wheel changes The load is then always concentrated at the lower part of the inner wheel and the function of the inner wheel is to support the frame only at this point This concept is featured in GBO Design Engineering s concept of a diwheel GBO 2005 as shown in Figure 2 9 Figure 2 9 A pedal powered diwheel concept vehicle GBO 2005 Another design concept used for the inner wheel is a spoked wheel which then attaches to an axle and bypasses an inner wheel to attach to the space frame This design does not need idler wheels to guide the outer wheel instead it uses a large hub and axle The inner frame then hangs and pivots from the axle 2 1 4 Idler Wheel The function of the idler wheels is to support and guide the outer wheel The idlers make contact with the outer wheel at certain points along its profile This contact has been solved through a variety of solutions THE TRINITY Figure 2 5 uses custom idler rollers made from nylon with inset bearings Southall 2009 This solution is also used in the ROCKET ROADSTER by Kerry McLean Figure 2 4 The Vereycken diwheel uses a similar idler wheel with a square inner profile Figure 2 10 Another solution for guiding idler wheels
200. s study of the integrity of the space frame design was performed Section 6 2 The final concept design of the space frame uses tubular framing and is shown in Figure 6 6 The seat and harness selected in Chapter 4 have been incorporated into this design The harness is bolted to two rails under the seat and the harness straps are fed through holes in the seat and are looped around tubing of the frame The harness is also connected to a gusset plate which strengthens the frame 52 6 1 Space Frame Design using two eye bolts The brake lever is mounted to a tubular bar extending from the main space frame The battery and electronics box are mounted underneath the passenger s legs and behind the seat respectively The overall design of the space frame consists of a ladder frame which begins at the feet of the occupant and extends around the seat to a roll cage protecting the occupant s head The framing running between the wheels of the diwheel are bent to meet the inner wheel at connecting brackets The brackets weld to the space frame and bolt to the inner wheel enabling the inner wheel and space frame to be separated Figure 6 6 Final space frame design concept 6 1 4 Electronics Box An electronics box is incorporated in to the space frame for the housing and protection of the control electronics of the diwheel The final design shown in Figure 6 7 is a folded Aluminium sheet metal box with a lid that pivots open A removable electronics moun
201. se train PT6 or PT1 digital value for the same encoder is sensed A value of zero indicates the motor is rotating clockwise and a value of one indicates it is rotating anticlockwise These values are used to estimate the speed of the motor This process is repeated for each encoder in succession as shown in Figure 8 2 This method worked effectively for most frequencies however an adaptive sampling interval should be implemented to gain better results for all speeds of the motor This method adjusts the sampling period Tel to match the last pulse count Lygouras 2000 and can improve the accuracy of the speed estimates Because the constant time elapsed method uses interrupts these interrupts should be given a lower priority to the other interrupts which run the control loops 8 5 Measuring Inner Frame Angle and Slosh Rate This section describes the integration of sensors to measure the inner frame s angle with respect to the gravity vector and its angular velocity slosh rate These measurements are required for the slosh and inversion control systems Two sensors are located on EDWARD to provide measurements of these two states 0 6 a two axis accelerometer and a single axis gyrosensor Figure 8 3 109 8 Software Design Read sensors Scale and remove joystick center offset Is X Y axis less than 5 deadzone Scale up X Y axis X Y axis is zero Drive Button Pressed AND Br
202. sed here was simply to feed back the angular rate of the inner frame 6 This decision was based on simplicity and availability of a state measure ment from a solid state gyroscope The final controller was u Vm 0 0 31 0 x 310 32 and was chosen to make the poles of the linearised dy namics entirely real Note that the positive sign for this term arises from the fact that a positive motor torque leads to a negative acceleration of the inner frame see Equation 19 3 2 Swing up control The swing up controllers developed for other under actuated non linear planar mechanical systems such as the inverted pendulum are applicable to the swing up of the diwheel since their dynamics are similar As trom and Furuta 1996 Yoshida 1999 Wang and Fang 2004 Almost all early works on swing up controllers used a bang bang switching approach to drive the po tential energy of the pendulum in this case the in ner frame to the inverted state More recently fuzzy control has been used to swing up and balance in verted pendulums Martynenko and Formal skii 2005 Chang et al 2007 In this paper two approaches will be investigated a simple positive velocity feedback con troller and a fuzzy controller Positive velocity feedback One very simple strategy to swing the inner frame to the upright position is to simply move the complex poles from the left hand of the s plane to the right hand side by feeding back a positive velo
203. sible Major High Incidents may happen if people are O El The Diwheel is only operated at N NA movement of the plant eg around the diwheel while it is running O Su open space where no O Tipping First aid or medical treatment may be unauthorized people are close i required L En by L Falling O ls E Rolling over E XX Ad q Rolling forward PPE C The plant s load C Under between plant and a structure eg wall C Inability to apply brake C Falling off the plant C Part of the plant collapsing changing shape C Other issues C No O NA Can anyone be cut stabbed or punctured amputated by coming into contact with Moving plant or parts Possible First Aid Medium Only a few sharp edges present and O El Protective guards are designed N NA C Sharp or flying objects protective guards are installed Unlikely O Su and applied on idler wheel pinch El Workplaces elected to cause any major injuries BJ En points C Work pieces disintegrating O ls C Other O Ad issues C PPE C No LJ NA Page 4 of 10 Project No 757 Issue No 1 Ss THE UNIVERSITY F ADELAIDE E Ei PROJECT RISK ASSESSMENT Hazard Identified Can anyone be injured from an electrical shock Likelihood Consequence Score Comments i e when and where hazard is present task activity Hierarchy of Control Current Controls Action Tfr to Required CAR YIN YIN C Water near equipment NA NA NA Po
204. sing Bryson s rule which suggests diagonal matrices such that Qu 1 max acceptable value of 2 R 1 max acceptable value of u The states 0 and 6 as well as the control input were restricted to 15 degrees 2100RPM and 36 volts respectively q 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 gq q 00 se 7 54 00 4 q 0 0 3P me and the effort penalty was 1 2 Ral i E R with cost function i 2 A 2 Solving the continuous algebraic Riccati equation returns the optimal control gains k 190 0 45 13 7 57 It is important to note that the stabilisation of the inner frame in the inverted position does not require a measurement for y which is important if this state is not measured 103 7 Control System Design 7 6 3 Combining the Swing up and Balancing Controller The SIMULINK block diagram of the combined swing up and balancing controller is shown in Figure 7 19 If the Control Mode is set to 3 the second multiport switch feeds the third input through to the Control Voltage output The system states 2DoF are fed into a Fuzzy Logic Controller as well as a block which determines if 0 is within 10 degrees of the inverted position If 0 is less than 10 degrees the first multiport switch feeds the second input through to the second multiport switch which consists of the balancing control gains multiplied by the states K xu If 0 is not within 10 degrees of the inverted position the Fuzzy Logic Controller
205. sponse for 1V the increase in amplitude from 0 025 rad to 1 2 rad 121 9 Experimental Results and System Integration Open loop step response for 6 30V 0 5 0 T amp 0 5 1 1 1 2 3 4 5 6 7 8 9 10 Time seconds Open loop step response for d8 dt 30V 4 T T T T T T 22 3 s 0 o 2 f f f 0 1 2 3 4 5 6 7 8 9 10 Time seconds Figure 9 2 Open loop step response with 60kg driver at 30 volts predicts that the voltage input in the test was 1 2 0 025 48V This is impossible as the full voltage available during testing was 36V The discrepancy may be caused an underestimation of the differential damping in the model This increase in damping friction has the following result e The speed of the motor decreases reducing the back emf e More current flows in the motor due to the greater armature voltage increasing motor torque e The increased torque causes a larger reaction torque due to gravity therefore increasing 0 Therefore according to this test the model s differential damping bj is closer to big 30 48Nm s rad oe IN 0 025 x 30 assuming a voltage of 30V was applied 122 9 2 System Tests Closed loop step response for 6 Kp 31 0 5 T T T T T T T T T T SA A 7 T 0 5 J Do SFE 4 1 5 i 0 1 2 3 4 5 6 7 8 9 10 Time seconds Closed loop step response for 6 Kp 31 1 T T T T T T T T T T
206. steel Damper fitting plates Spring dampers set Idlers Idler guards Idler guard brushes Drive wheel Sprocket and chain Frame Seat Harness Quick lock clamps for joystick mount Brake caliper Brake lever Brake cable Bearings Paint Dragon board Battery SLA GEL Gyro Accelerometer Encoders Motor Includes both sets Joystick Motor controller Roboteq Circuit boards cables components Total Cost Table 10 2 University of Adelaide workshop hours Total Hours Mechanical Workshop 304 Electrical Workshop 98 Total Workshop Hours 402 200 583 5 110 10000 385 108 80 120 72 400 170 50 30 0 25 9 12 8 210 99 200 360 89 98 29 89 60 1100 39 95 850 384 01 4720 03 10 2 Cost Analysis Table 10 3 Student work hours C Dyer E Schumann J Harvey K Fulton T Zhu March 19 54 46 41 69 April 40 82 30 37 72 May 62 113 75 67 143 June 30 55 30 4 61 July 41 27 31 29 30 August 62 35 5 31 45 60 September 47 37 5 67 50 49 October 78 98 87 71 68 Student Hours 379 502 437 344 552 Team Total Hours 2174 135 10 Final Design Analysis 136 11 Summary 11 1 Future Work There are several modifications that can be implemented to improve the performance and functionality of EDWARD These were not implemented due to the limited time frame caused by project delays e Implementing closed loop speed and yaw control allowing a more intuitive dr
207. stom pneumatic solution which provided significant vibration isolation Figure 2 8 Figure 2 8 shows the wheel using a custom tubeless tyre and rim arrangement modified with a flange for rolling contact with the supporting idler wheels The Swedish diwheel does not use any tyre and the steel wheels make direct contact with the road The mini monowheel of laFrance Bressen Figure 2 7 uses a tractor tyre modified for use in the monowheel design With the use of existing wheels the tyre is already integrated into the wheel and requires no further customisation for use in a diwheel Larger tractor wheels and tyres have been used in the design of several monowheels shown on the Museum of Retro Technology website Monowheels 2009 me Figure 2 8 Construction of the patent of Vereycken Unknown 1935 2 1 3 Inner Wheel The inner wheel see Figure 2 1 is a rigid structure supporting the guidance and drive mecha nisms of the outer wheel and connects to the main frame Typical monowheel designs require 2 Background this system to accommodate the seating of the passenger as featured in the monowheels shown in Figures 2 4 and 2 7 However a diwheel does not require the seating of the passenger in either inner wheel The inner wheel design of a diwheel then has an additional function of sup porting an inner frame of the vehicle where the passenger is seated The design of a monowheel often uses a circular rolled inner wheel see Figure 2 4 which i
208. t input and the state and input matrices are given by A A 0 0 ap Ji Jo 0 das 1 n 0 0 0 an Jid Gn dde Jiag 0 ian ar b z A ar b12 arb agar 0 Jz ar b 2 J2 ar b z J2b1 and 0 1 0 aR Ji Jo J an Ja ar The poles of linearised dynamics in the lower equilibrium position are at s 0 0 22 2 591 0 26 with the complex poles having a damping ratio of 0 083 The transfer function from torque 7 to slosh angle 0 exhibits one zero on the origin which is expected as at the steady state 0 s 0 0 It is interesting to note that the transfer function from 7 to outer wheel angle y exhibits two purely imaginary zeros at s 3 44i The presence of purely imaginary zeros is similar to that found in other systems exhibiting slosh such as the ball and hoop system Wellstead and Readman The implication is that if the motor is driven with a sinusoidal input at the frequency of the zeros then the wheel will stand still and only the inner frame will move Note that for the case of a voltage input u Vm then the damping term arising from the differential velocity of the frame and wheel increases from bi b12 bm resulting in open loop poles at s 0 0 34 2 55 0 37 and the state input matrix B needs to be multiplied by A Linearising about upright inverted position Linearising the non linear dynamics given by Equations 7 24 and 7 26 about the upright pos
209. tal translation of the diwheel centre from an earth centred reference frame Both the rotation of the outer wheels and the inner assembly are taken from a zero datum given by the positive y axis As the horizontal translation of the diwheel is dependent on the rotation of the outer wheels the system can be classified as having two degrees of freedom 7 1 2 Euler lagrange Derivation The Euler Lagrange approach follows this basic process 74 7 1 Analytical Derivation of the System Dynamics 1 Define the potential and kinetic energies of each body of the system in terms of its coordinates 9 and y in the diwheel s case 2 Find the Lagrangian L given by the difference in the kinetic and potential energies of the system L E Ep 7 1 3 Apply the Euler Lagrange Equations given by d ES ia a dt 0d Odi where q are generalised coordinates and F are generalised forces This yields an expression describing the systems dynamics of the form M q q C q 4 G q F 7 3 Velocities To define the kinetic energy of the system the rotational and translational velocities of each body are found The rotational velocities and are the angular velocities of the outer wheels and the inner frame assembly respectively The translational velocity of the centre of gravity CoG of the outer wheels body 1 in the z direction is Vie R 7 4 and there is no translation in the y direction for body 1 The translatio
210. tation Orientating the accelerometer with one axis horizontal however meant that only measurements of the remaining two axes were required to measure the slosh angle The gyrosensor measures angular rates about one axis only and so this axis must be aligned parallel to that of the rotational axis of the inner frame axis through the centre of the outer wheels The two sensors were mounted on the sensor board in a fashion such that the gyrosensor axis was parallel to one of the accelerometer axes resulting in only two accelerometer axis readings being required To align the gyrosensor axis with the inner frame s axis of rotation the board required mounting on a surface that would always remain in the zy plane The inner wall of the electronics box was chosen as a suitable mounting point Figure 5 10 Picture of the sensor board with the gyrosensor and accelerometer 45 5 Electrical Hardware 5 7 Encoder Sensors The velocity measurement was required to provide speed and yaw feedback with direction to the closed loop control systems Typical devices measure the angular velocity of a rotating part which may then be converted into a desired system velocity measurement using proper scaling Accessible rotating parts of the system included the motor sprocket the chain and the brake s disc A number of speed measurement devices were considered including tachometers resolvers and encoders A tachometer made from a brushed DC motor was dismisse
211. the ar mature current is given by T Nn Kyi 23 Combining Equations 22 and 23 gives the differen tial torque in terms of applied voltage T Nn Km Vn NnsKm Rm 24 Inserting Equation 24 into Equations 17 and 18 yields the differential equations of the fully coupled electro mechanical system Solution to the Differential Equations of the Coupled Electro Mechanical System Equations 19 and 21 may be rewritten in terms of an input voltage to the motors by substituting Equation 24 to give i 1 ae arcos 0 bip i aR cos 0 1 Nn Km vn b12 bm 8 2 a sin 0 cos 0 6 Ja sin 0 25 where bm NnsKm is the effective damping from the back EMF and SS l AE 2 p Ph ie apr cos 0 a x Ro arsin 6 6 apa sin 9 cos 0 26 2 3 Linearised Dynamics The dynamics of the plant have been linearised about two operating conditions the downward stable posi tion and the upright unstable position Linearising about downward position Using a Jacobian linearisation the non linear dynamics given by Equations 19 and 21 about the downward position 0 6 p y 0 may be approximated by the linear state equations x Ax Bu 27 ame ae fo de where x 0 p080 is the state vector u T is the plant input and the state and input matrices are given by a 1 aR a Jide 0 0 aR idh 0 s 0 0 0 ar Ji
212. the range of operation This range is usually specified in term of the gravitational acceleration with common full scale ranges of 1 5g 2g 4g and 6g Although accelerometers with greater range than this are avail able they suffer reduced resolution For the diwheel project it was deemed that anything more than 6g was quite excessive Sourcing the accelerometer from SparkFun with an aim to minimise cost resulted in the the accelerometer shown in Figure 4 15 already mounted on a breakout board The accelerometer measures changes both static and dynamic in the acceleration of 3 mutually perpendicular axes and has selectable full scale ranges making it very versatile Setting the full scale range to 1 58 provides the highest sensitivity but 2g has been chosen as the sensitivity The accelerometer has the following specifications e Variable full scale range 1 5g 2g 4g and 6g e Triple axis measurement 34 4 12 Steering Control e Operating voltage 2 2V 3 6V e Sensitivity 860 185mV g depending on range selection Figure 4 15 Accelerometer Breakout board Sparkfun com 2009 4 12 Steering Control Steering of the diwheel can be easily accomplished using a number of techniques The design of a diwheel from an electrical stand point makes steering and control more flexible particularly because steering can be completely controlled by signals arising from a microcontroller Several steering methods were considered by Dyer et al 20
213. the translational DoF is along the x axis An initial condition the IC block has been included so that an initial translational velocity rotation velocity or rotational angle of the inner frame may be set at the start of a simulation Two sensors are connected to the Body DOF block each of which measures the kinematic state 82 7 2 Simmechanics Derivation of the System Dynamics Tal merno IS This model simulates the system using SimMechanics for the governing DEs input votage 9 Motor Current Amps x de d2x m m s m s 2 Voltage V Drivewheel Angular Velocity rad s th dth d2th rad rad s rad s 2 2 Disturbance Torque Drivewheel Angle rad Drivewheel Angular Velocity To Motor Disturbance Torque Nm Outer wheel torque Nm p Diwheel Inner ring Ab To Inner Ring To Outer Wheel Drivewheel with gearing and viscous friction Figure 7 2 The main SIMMECHANICS block diagram of each DoF These measurements can be viewed on Scopes and exported to MATLAB for analysis 7 2 3 Drive System The SIMMECHANICS model of the drive system shown in Figure 7 4 models the behaviour of the friction drive wheel as a joint between the inner frame and the outer wheels In MATLAB s SIMMECHANICS toolbox forces and torques may only be applied through joints
214. the wheels to achieve steering and remarkable manoeuvrability Serious wheels 2005 This allows swift and complete yawing of the vehicle while stationary The control of the diwheel structure is influenced by its dynamics which are mainly con strained by it s geometry and the distribution of mass Rotational motion of the diwheel s inner frame relative to the earth centred frame is termed slosh and a complete revolution is termed tumbling or gerbiling The designs of many monowheels avoid this behaviour by having a centre of mass far from the centre of rotation This can be achieved by adding mass to the in ner frame as was done by Kerry Mclean Parks 2005 or by designing the seating position and engine mount far from the centre of rotation However if intended gerbiling can be facilitated in a diwheel by keeping the centre of mass close to the central axis 2 1 7 Brakes Mechanical braking systems aim to remove rotational and translational kinetic energy from the vehicle In the case of the diwheel this action may also convert translational energy into rotational energy of the inner ring which may resulting in gerbiling Mechanical braking systems need to be mounted to the drive wheel axle or the outer wheel Often the drive wheel is acted on directly Cable drawn disc brakes are used in the design of The Trinity Figure 2 5 mounted directly to the hub of the drive wheel The Vereycken diwheel Figure 2 8 also uses brake pads which are
215. thematical model used to simulate the diwheel s sloshing behaviour a number of tests were performed Data for 6 from the gyrosensor was recorded at approximately 50Hz for 10 seconds producing 500 data points After the data had been recorded it was sent to MATLAB using a function accessible from the menu on the Dragonboard Though functional this method of logging is not particu larly convenient and in future the system could be setup to log all data to the SD card in the desired format and send or transfer it to a computer in one go Both open loop and closed loop testing was performed to give an indication of the suitability of slosh control on the platform 120 9 2 System Tests Figure 9 1 Damping tests 9 2 1 Open loop slosh The first slosh control test was an open loop step response applying approximately 24V to both motors for 10 seconds with a 60kg driver Figure 9 2 shows the response for 6 and 6 Because the accelerometer was damaged at this stage the estimate for the inner frame angle has been calculated using numerical integration of the slosh angular rate 0 The bias of the slosh angular rate was noted as 0 3326 rad s and was accounted for before the integration was performed Although the bias of the gyrosensor will increase with time this constant approximation is justified as the bias increases slowly over time Unfortunately during testing data logging of the control signal was overlooked and has not been recorded
216. ther The H bridge controller topology was chosen for its 1 Simplicity and robust design 2 Varied and simple switching schemes 3 Four quadrant 4QD operation and was to be controlled through Pulse Width Modulation PWM of the four switches ac cording to a particular switching scheme Dyer et al 2009 discuss in detail the numerous switching schemes available to the user when switching a H bridge which include e Sign magnitude e Locked antiphase e Active field collapse e Modified active field collapse e Synchron collapse After considering the normal operation of the diwheel the Sign magnitude drive was deemed the most suitable switching option as it is commonly used when the direction of the motor does not change frequently Using this drive scheme regenerative braking was an option and Dyer et al 2009 describe its implementation However as discussed in Section 5 1 the final motor controllers did not allow access to the transistors in the H bridge instead being controlled by one of three methods Analog R C control or Serial control 30 4 8 Batteries 4 8 Batteries The selection of the motors was the main driver for the choice of batteries to be used in the diwheel The batteries needed to provide adequate voltage for the required motor speeds and supply sufficient current to achieve adequate torque at high loads Several types of batteries were considered for use and details of the advantages and disadvantage
217. tially estimate the required eccentricity of the center of gravity a simplistic energy approach was taken where the diwheel s kinetic energy traveling at 8km hr is converted into potential energy in a braking situation to invert the diwheel This calculation can be found in Appendix B 3 and gives an initial design value for the eccentricity of the center of gravity of the diwheel of 0 29 metres The dynamics driving the design of the width between wheels of the diwheel is a worse case yawing situation If one wheel locks and the diwheel yaws about one wheel then a centrifugal 50 6 1 Space Frame Design Requirement Dynamics Requirement Requirement Diwheel width Centre of gravity Ergonomics Inner wheel gt Requirement Passenger Component Component Component Component Component Batteries Seat amp Harness Brake lever Joystick Footrest Electronics Component Lal e Space frame Figure 6 3 Functional diagram of the space frame Figure 6 4 Assembly of the final concept design of the space frame force created from this rotation must balance with the weight of the diwheel to keep both wheels on the ground The calculation for this can be found in Appendix B 3 and gives an approximate initial width of 1 2 metres 5l 6 Mechanical Design 6 1 2 Ergonomics Due to the variability of the human form a system to be use
218. ting plate is fastened in the box with spacers in between The electronics box is ventilated by incorporating a 12 Volt DC 60mm diameter computer fan placed on the inside of a louvered panel of the box Figure 6 7 Final concept of the electronics box 93 6 Mechanical Design 6 1 5 Battery Box The batteries chosen represent a significant proportion of the weight of the diwheel and must be supported accordingly These batteries are mounted below the footrest in an open box constructed of folded sheet steel which is bolted to the space frame structure The battery enclosure has a lid fastened by wing nuts to allow easy access to the batteries The battery terminals are exposed when the lid is secured so that they can be connected and charged directly Electronics box Mounting bracket Figure 6 8 Final concept of the battery box 6 1 6 Joystick Mounting The joystick used for the diwheel s user control input needs to be mounted rigidly in an er gonomic location The relative frailness of the joystick s structure prevents it from taking any loads and the harness should deter the occupant from loading the joystick To prevent any overloading of the joystick it was designed that it be mounted on a pivoting structure which provides support for the joystick in its intended operational location and retracts when the occupant enters and exits the seat or when a significant load is placed on the joystick The final conce
219. tor Figure 7 18 shows the simplest block diagram used for the controller It consists of a full state feedback regulator which ensures all closed loop poles are stable at the inverted position This means that any sufficiently bounded impulse exerted on the inner frame when inverted will produce a bounded output for all of the system s states The system model contains the A matrix which determines how the state vector evolves due to the internal dynamics of the system and the B matrix which determines how the states are affected by the actuators Because there is only one input to the system u is a scalar The control input consists of the motor torque which is applied differentially to the inner frame and outer wheels The C matrix maps the system states to outputs measured by sensors Linear Quadratic Regulator The feedback gains for relocating the closed loop poles have been determined using the linear quadratic regulator LQR approach This approach has been used successfully in the inverted pendulum problem and its many variants Chang et al 2007 Lauwers et al 2006 Muskinja and 102 7 6 Inversion Control Figure 7 18 Basic State feedback regulator Tovornik 2006 Determination of the required feedback gains K requires the minimisation of a performance index or cost function J x7 r Qx r u r Ru r dt 7 53 Such a solution for the matrices Q and R is termed optimal The matrices Q and R were determined u
220. tor controller contains internal circuitry which converts speed torque commands from some simple setpoint such as a resistor into the appropriate output The alternative is a motor controller which derives control signals from the microcontroller itself such a motor controller acts as a power amplifier Generally the intelligent controller is slightly inflexible as it is usually tailored to a particular application This project requires a good deal of flexiblility in terms of the motor controller since it has several proposed modes of operation e Free running mode e Slosh control mode e Inverted mode In the free running mode it is more intuitive for the user to control speed rather than accelera tion while inverted mode may require torque control The voltage applied to a DC motor with no load determines its speed while the current controls the motor s torque Hughes 2006 Dyer et al 2009 present two controller possibilities a contactor controller and H bridge controller The decision for the motor controller was motivated by the following factors e Cost e Power handling capability e Ease of control e Performance with regard to control goals and resulted in the design of a H bridge motor controller adapted from the OPEN SOURCE MOTOR CONTROLLER H bridge design The motor controller board is shown in Figure 5 2 This controller was designed for 4 quadrant operation as was specified in the project goals Unfortunately during
221. tress occurs in the steel Tmar 88 0MPa e Maximum shear stress in the idler wheel Tmaz 52 6MPa e Maximum contact pressure pg 97 2MPa e Hertzian contact theory results predict a maximum shear in material with the lower stiffness Nylon Tmax 33 3MPa e Hertzian contact theory predicts a maximum contact pressure pg 104 1MPa The conclusions that may be deduced from these results are based on the maximum shear stress criterion comparing the resultant shear stress to half that of the material s yield stress 0 Tmax lt TmaxAllow 6 1 The yield strength of steel used in this comparison is og 250MPa and the yield strength of Nylon 6 6 as given in 180527 is oy 60 90MPa Nyl 2009 As Stainless Steel was the actual material used for the outer wheels these results were only intended to be used as a guide in the design process Furthermore the effects of work hardening on the outer wheel during the rolling process have not been considered in this section Finally the following points are important e Local high contact stresses indicate that the contact zones of the idler and outer wheel are expected to deform and wear during operation 56 6 2 Finite Element Analysis Maximum shear stress in outer wheel 88 0MPa Idler wheel surface Outer wheel surface Maximum shear stress in idler 52 6MPa Figure 6 11 Shear stress contour plot of the idler lower body and outer wheel upper body contact
222. uid fueled rockets ships and tankers Aboel Hassan et al 2009 Readman and Wellstead accessed Aug 2009 Wellstead accessed Aug 2009 Any number of suitable linear and non linear control strategies can be used to suppress the rocking sloshing motion of the diwheel Readman and Well stead accessed Aug 2009 fed back the slosh angle of a ball in a hoop equivalent to in the diwheel to re strict the slosh which was equivalent to increasing the torque arising from the offset in the CoG of the inner Table 1 Parameters used to define the model Note that the terms for the wheels and motors account for both acting together Part Parameter Value Wheels mi 50 3 kg J 26 1 kg m Frame ma 218 kg Jo 48 4 kg m Lengths R 720 mm Rix R 720 mm r 140 mm e 160 mm Damping bye 30 Nm s rad by 12 Nm s rad Motor Veat 48 V Rm 0 314 Ohms Lm 0 3 mHenry Km 65 mNm A Transmission N 4 5 14 Ns 7 frame It was found that this technique is effective as it drives two complex closed loop poles towards the plant zeros An alternative and obvious solution is to increase damping to reduce slosh using velocity feedback An other common technique is input also known as com mand shaping which involves modifying the reference command by convolving it with a set of self destructive impulses that act against the complex poles in the plant This approach is effectively pole zero cancellation and is not robust The approach u
223. uid slosh was considerably suppressed for an oscillating excitation of the tank Gandhi et al 2009 This literature provides constructive insight into the types of control methods to use in the suppression of slosh in the diwheel 2 3 3 Application to the Diwheel All relevant literature reviewed for the project regarding slosh and slosh control used the La grangian method and the Euler Lagrange equations to determine the dynamics of the system This method is straightforward and relatively simple to perform making it the preferred method for the derivation of the diwheel s dynamics It is probable that the slosh control to be implemented on the diwheel will consist of a form of PID control in keeping with the methods used for the ball and hoop system and 2DoF slosh rig The results of the Ball and Hoop control system suggests that a purely proportional controller feeding back the slosh angle may be sufficient in providing satisfactory control and rider comfort Should purely proportional control not provide sufficient results a Proportional Derivative PD controller controller may be necessary 2 4 Yaw Control Yaw control regulates the rate at which the diwheel turns about its vertical axis The imple mentation of yaw control will reduce the undesired yawing of the diwheel whilst it is travelling in a straight line and will also allow the turning of the vehicle to be smooth and responsive Re cent research on closed loop yaw control has
224. ulated supply for the Dragonboard e 5V regulated supply for the gyrosensor encoder sensors e 3 3V regulated supply for the accelerometer e 12V regulated supply for the current sensor e 36 48V connections for each motor Power at 36V is fed directly from the batteries through two high current horn relays to the motor controllers These relays are controlled by the Power Control Board described in Section 5 3 The single ROBOTEQ motor controller is connected to the series connection of three batteries providing 36V through the high current relays The Power Board also takes an unregulated 24V input from the series connection of two batteries and generates the regulated system voltages Figure 5 4 Power board use to provide regulated voltages to system components 40 5 3 Power Control Board 5 3 Power Control Board To increase safety of the system and add redundnacy a power control board has been added to give the driver control over the power to the motors This power control board implements an emergency stop button and a start button to form an emergency shut off circuit The emergency shut off circuit consists of a mushroom button which opens two relays when pressed cutting power to both motors It is not essential that power be removed from the power distribution board which may continue to operate and so the stop circuit only cuts power to the motor controller allowing the Dragonboard to continue running In this situtation a si
225. utton in preparation for next Vehicle use Machine Pre operational Safety Checks Safety Precautions that MUST be Observed Visual inspection of machine to verify it is in good operational order ensuring no damage to any stationary or moving parts electrical cords etc Any unsafe equipment is to be reported to project supervisor and other project members Ensure that the 5 point safety belt harness is worn tightly WARNING FAILURE TO PROPERLY APPLY SAFETY BELT MAY RESULT IN DEATH OR CRITICAL INJURY Be aware of other activities happening in the immediate area Ensure that no slip and or trip hazards are present Ensure that NO unauthorised personnel are close to the vehicle Ask project staff to switch on the vehicle Check that the Emergency Stop is working Check that all machine guards electrical interlocks and Emergency Stop micro switch are correctly positioned locked in place and in proper working condition IF IN DOUBT ASK Ensure O O O O O Knowledge of the location of Emergency Stop Button Driver is aware of all functions of the joystick and controller buttons Driver is aware of the consequences of operating this vehicle at excessive speed Driver is aware of the consequences of making sharp turns at high speed Driver is aware of the consequences of locking the braking system Vehicle Operation 0 D D D D D DO D DO DU CO Do not switch on the main power before driver is seated and safely fastene
226. wer source is only 12V batteries O El NA N NA E Plant located near or in EJ Su contact with exposed live O En electrical conductors ee O ls C Leads switch in poor condition Ad C Overhead and underground PPE wires C Other issues x No O NA Can anyone be injured by an explosion L Gas O El C Vapour O Su Dust C En L Liquid DJ Is C Other issues Dl Ad O No O PPE x NA Can anyone be burnt due to Friction C Contact with moving parts or O El surface of the plant O Su C Material handled by the plant O En C Other issues O ls No O Ad J NA O PPE Page 5 of 10 Project No 757 Issue No 1 Ss THE UNIVERSITY F ADELAIDE E Ei PROJECT RISK ASSESSMENT Hazard Identified Likelihood Consequence Score Comments i e when and where hazard is present task activity Hierarchy of Control Current Controls Action Required YIN Tfr to CAR YIN Can anyone be struck by moving objects due to L Plant materials being ejected Unlikely Minor Low All components are moving at low O El No contact is allowed with the N NA XX Plant material movement speed No serious injury will occur O Su running diwheel except the E iher S iss E En operator while it is running O No O ls O NA Xx Ad Ol PPE C Lack of oxygen O El C Atmospheric contamination C su C Enguifment O En C Other issues O ls C No O Ad xX NA C PPE C High low temperature O El C Naked
227. xii 1 Introduction A diwheel is a vehicle which consists of a two large outer wheels that completely encompass an inner frame The inner frame is free to rotate within the wheels and is typically supported by a common axle or idler wheels which roll on the inner surface of the wheels The outer wheels are driven from the inner frame and forward motion is achieved through a reaction torque generated by the eccentricity of the centre of gravity of the inner frame Diwheels like their more popular cousins the monowheel have been around for almost one and a half centuries Cardini 2006 Each of these platforms suffer from two common issues affecting driver comfort those of slosh and tumbling also known as gerbiling Slosh consist of the oscillation of the inner frame and it occurs in all monowheels and diwheels where the centre of gravity of the inner frame is offset from the centre line of the wheels It is very noticable as these platforms typically have low damping between the wheel and the frame in order to minimise power consumption during locomotion In addition during severe braking or acceleration the inner frame may tumble relative to the earth centred frame which affects the ability of the driver to control the platform The aim of the project was to design and build an electric diwheel and to implement several control strategies to modify its dynamics The first was a slosh controller with the pur pose of minimising the rocking motion
228. y consist of a pneumatic solution or solid foam filled solution A custom pneumatic or solid foam filled solution for the outer wheel tyre of the diwheel proved infeasible These alternative solutions were dismissed early in the design process as they were deemed to have been too expensive and problematic to implement Suspension system AT There shall be some means of isolating the driver from the vibration levels experienced at the wheel The final diwheel prototype exhibits a working suspension system which uses automotive spring dampers to allow movement in the idler wheels This system provides significant movement of the inner frame relative to the outer wheels and maintains full idler wheel contact with the outer wheels despite undulations in the circularity of the outer wheels This reduces the vibration levels transmitted to the driver and provides good driver comfort indicating the ac complishment of this goal Real time Radio Control BT The di wheel shall be capable of Remote Control via an RF link This shall enable command control and simulation of user input into the system Remote control of the diwheel was not achieved However the electronic hardware in place is conducive to the implementation of a RF link to control the diwheel remotely This goal was not given high priority with respect to other goals and as such was not deemed to be feasible to implement within time constraints 132 10 2 Cost Analys
229. zzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TITLE Damper fitting plates top sides and bottom TOLERANCE ANGLES 0 1 LENGTH 0 1 UNLESS OTHERWISE SPECIFIED SCALE 1 15 QUANTITY top 4 sides 8 bottom 4 Quantity 4 Please make the 37mm hole for interference fit for 37mm OD and 12mm ID deep grove ball bearings bearing supplied Chamfer with 1 mm setback DRAWN 05 26 09 CHECKED Ben Cazzolato DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 DETAIL A TOLERANCE ANGLES 20 7 LENGTH 204 PO Ette tearing housing UNLESS OTHERWISE SPECIFIED SCALE 1 11 QUANTITY 4 Drive swing arm part 1 Drive swing arm part 2 drive swing arm part 1 weld around joints drive swing X1 arm part 2 weld around both joints Joints subject to cyclic load e weld another Drive swing arm assembly with the drive swing arm part 2 slot on the side Shown in the graph below AA Quantity 1 for slot on the left 1 for slot on the right DIWHEEL WHEELBIKE SUPERVISOR Ben Cazzolato CONTACT 0413647219 TOLERANCE ANGLES 0 LENGTH 201 TITLE Drive swing arm assembly UNLESS OTHERWISE SPECIFIED SCALE 1 0 25 QUANTITY 1 for Left and 1 for right make this item with 35mm side 3mm thickness steel SHS Quantity 4 DRAWN 05 27 09 CHECKED Ben Cazzolato DIWHEEL WHEE

The LyX User's Guide - University of Adelaide

Contents

Download Pdf Manuals

Related Search

Related Contents