Troubleshooting ECU Programmed by Bodybuilders
Contents
1. Troubleshooting ECU Programmed by Bodybuilders Tony Lindgren Department of Computer and System Sciences Stockholm University Forum 100 164 40 Kista Sweden and Scania CV AB Service Support Solutions YSNS Verkstadsvdgen 17 By 280 151 87 S dert lje Sweden Abstract Having an Electronic Control Unit ECU which is programmable by external parties puts new requirements on troubleshooting In this paper we describe how one could solve the problems of both troubleshooting additional equipment added by bodybuilders and facilitating their need to use signals from vehicles in an easy way in order to interact with their additional equipment In this paper we look at bodybuilder s additional equipment for heavy trucks but our technique for troubleshooting should be equally relevant for other applications with similar conditions Keywords Diagnostics Simulation Reconfigurable hardware Signal processing systems I INTRODUCTION In heavy truck industry bodybuilders use the trucks as platforms for their additional hardware There exists a wide variety of additional hardware ranging from concrete mixers to cranes Other industries face similar type of add ons to the base platform to suit the needs of the user Applications that adhere to this general description are ranging from military jet planes to farming equipment They have the same basic demand in that the user want to equip a base platform to suit their needs and want this e2. and that the driver is applying the brakes if this is so the vehicle kneels and it is possible to open the vehicles doors A simulation view is available for the user to validate their logical expression The user can in this view set the values of input and output ports to either 7 true O false or Where the question mark denotes a free unbound variable the possible assignments give rise to different scenarios It could also be the case that the user put the system in a state that is not valid for example having a branching operator connected to two output ports where they are true respectively false In figure 2 the simulation view is shown as can be seen when all input and output are question marks four 4 different scenarios are possible Each accessible via the tabs above the canvas a connection having a thick line illustrates a true value while a thin line illustrates a false value For example when the user has designed a logical ex pression and uses the simulation view to test the expression he might see that the tabs are empty This implies that the logical expression is faulty as the expression will never give rise to any activation of the additional equipment A truth table is also available for the user When the user is satisfied with the design of their logical expression they can load the program into the physical EDE III TROUBLESHOOTING The task that we ponder here has the following char acteristic
3. e software we wanted to be able to use different SAT solvers Hence to avoid getting stuck on one particular SAT solver we decided to use an internal format XML Schema Internal format Figure 3 Information layers of the module B Algorithm The algorithm is dependent on three 3 sources of infor mation the logical expression a wished state and a current state The logical expression is sent to the system in a XML format The wished state is defined as the state when a function is realized i e when the output of the logical expression for a particular function is interpreted to true The current state is the state that we observe upon the truck right now Hence when a workshop wants to troubleshoot a body builder s equipment the proposed diagnosis software is activated for the EDE The truck is set for activation of the bodybuilder s equipment The software now calculates the most probable reason of why the equipment is not functioning as intended The algorithm for the troubleshooting informally works as follows Given the current EDE get the logical expression and transform it into a SAT solver format Create a wished state for a particular bodybuilder functionality the wished state is usually found by setting the function s activation to 1 true and the rest of the involved ports values to unbound value Use this with the SAT solver to calculate all possible scenarios when this function can be realized i e a
4. er can define the logic of how their additional equipment should work in conjunction with the truck To the left the GUI has two tree structures showing the current hardware the EDE s connected which the bodybuilder are working with and the available CAN signals on the specific truck DAET elt Cone gt 2 Master hardware Function Activate doors Kneeling 2 Change Name F n Brake pedal depressed Kneeling 3 F Acceleration f ae n Vehicle speed above 5 kph Do La 3 Brake gt Parking brake 4 rivce brake Brake pedal de d tT Der cain ering Power tke ot Poner bain Tf protection k a aeS Vehicle speed Do 21 amp D Figure 1 Here we see the main window of a bodybuilder application In the figure only one 1 EDE is present by navigating in the structures it is possible to drag in pins of the EDE and signals to the main canvas of the GUI The same holds true for the operators at the bottom of the figure The user then create logical expression by connecting the input be it CAN information signals or pins to operators and from the operators to the output which can be pins or CAN control signals The example application has two major modes the on line mode and an off line mode When using the main window in off line mode the user needs to specify the hardware configuration i e the number of EDE s In addition to that the production date specification of a truck mus
5. es from this field have been used in a wide range of applications from electronic circuits to gas turbines 5 The problem setting we are looking at in this paper is easier than the typical problem formulation within this field Usually this involves a system description which typically is a description of components and their connections This system description together with observation are used when inferring usually through abduction or consistency which components that are faulty i e selecting one hypothesis out of all hypotheses that either explain or is consistent with the observations The parts that constitute model based diagnosis MBD problem formulation as described above are a system de scription SD and observations OBS upon the system together with the components COMPS of the system Be low is the MBD formulation for consistency and abduction diagnosis Consistency formulation SD U COMPS U OBS FL Abduction formulation SD U COMPS OBS In our setting we have the assumption that the internal components of our model cannot be faulty What can be faulty is objects either before our logical expression the input or after the logical expression the outputs This simplifies our troubleshooting of the system but also reduces our ability to isolate faults to these points i e to the input or output The rest of the article is organized as such in the next sec tion we will give an example o
6. f how a bodybuilder can add their logical functions to the EDE and how they can verify their design We will then look into how the workshop can troubleshoot the logical functions that control the additional equipment here we will also go into some of the technical details of how we realized this troubleshooting functionality In the section after we will look at the performance of the simulation algorithm and after that we will discuss our experiences using this technique so far and finally we will look at some related work II CREATING FUNCTIONS By using a computer program which can interact with the EDE that we want to program the bodybuilder can create functions that control their additional equipment Such a program could use a graphical interface GUI through which a bodybuilder can express their logical func tions Similar type of visual programming tools can be found for other programming tasks see 6 Functions can be formed using information signals from the vehicle as well as control signals for requesting functions from the vehicle The physical EDE also has a few input pins and output pins both digital and analog which can be used in the functions For bodybuilders with the need for many pins or very complex expressions it could be possible to connect a number of slave EDE s to the master EDE thus expanding the number of physical pins available In figure 1 an example GUI of such a program is shown here the bodybuild
7. he input for the third layer In table II the result of the simulation is shown The first column denote the number of layers the second column the number of input the third the number of operators the fourth the amount of time used by the CPU in milliseconds and the last column shows the amount of memory used in kilobytes Table II RESULTS No layers No inputs No operators CPU time Memory 1 4 3 0 2170 848 2 16 15 266 8721 824 3 64 63 3751 123643 440 Figure 5 shows how CPU time and Memory usage y axis correlate with the logical expressions complexity The CPU time is shown in milliseconds and the memory used in kilobytes In this case we use a crude measure of complexity expressed by number of inputs number of operators This gives the first layer complexity of 12 4 3 In this figure we can observe that the number of input re use affect the memory needs of the algorithm significantly Notice the sharp turn for the memory needs around 8700 kilobytes The reason for this is that the re used input does not create more variables in the constraint store of the SAT solver 14000 12000 gt Memory I cPv time Figure 5 The relation between complexity and CPU time and memory usage V RELATED WORK Mercedes have an ECU bodybuilder node called PSM which stands for Parameterable Special Module It is similar to our proposed EDE in that a bodybuilder can define log
8. ical expressions using a visual programming tool But to the authors knowledge they are lacking simulation and troubleshooting capabilities Iveco and DAF has on some trucks the possibility of bodybuilders to use CANOpen layer 1 for communica tion with their equipment But no information was found whether they have any troubleshooting assistance Volvo Scania and other manufacturers use a dedicated ECU that have certain signals that are available and the output on the ECU connectors are defined by adjusting parameters VI DISCUSSION AND CONCLUSION As we showed in the performance section even though we are using a modern SAT solver which uses techniques from 3 memory needs grows exponential with complexity although this is disturbing it has not been a practical limitation The reason for this is that the users tend not to create very large expressions instead the number of expressions can be large for some vehicles This does not affect the complexity of simulation or troubleshooting As noted in the section related work the way of creating logical expressions with a visual programming tool is not new for the truck industry but using these logical expressions to facilitating simulation and troubleshooting capabilities probably is which is the main contribution of this work The response from test users have been positive when using the prototype software They do see the benefit from using this tool as it will speed up their p
9. lder function The program displays these ports and hand over the rest to the workshop Hence further investigation might be needed to be able to point out which component s on the additional equipment and or truck is faulty The standard diagnostic services the formation of diag nostic trouble codes DTC guided diagnostics etc could of course still be used for troubleshooting the EDE as for other ECUs on a specific truck Our concern in this paper is how we support the workshop to answer the question why is the additional equipment no longer working properly A Architecture The troubleshooting has been implemented as a separate module We will briefly go through the different information layers and their responsibilities and motivate why the mod ule looks the way it does The module is set up as a server towards which another program can use the services exposed by the module The module offers two main services a simulation service and a troubleshooting service where the latter is an extension of the former The module deals with information in three 3 different formats as shown in figure 3 Firstly the format in which it communicates with outside world XML schema Secondly the internal format and thirdly the format of the satisfiability solver SAT used for more information about SAT solvers see 4 The reason of having this potentially not necessary internal format comes from the fact that when constructing th
10. ops to troubleshoot the bodybuilder s equipment If the workshop is lucky the truck operator had some form of electrical schematics of the additional equipment But it is not always the case and if more than one bodybuilder has added their equipment to the truck it could be the case that one bodybuilder alter another body builders electrical wiring to fit their needs Thus giving rise to the problem of not having up to date electrical schematics for additional equipment making the job at the workshop challenging and time consuming By expanding the capabilities of the DE used for interact ing with the bodybuilder s additional equipment it is possible to improve upon the above mentioned problems The term EDE will hence forward be used for DEs with expanded capabilities This includes creating tools that support body builders in realizing their logical functions that control their equipment within the EDE which eliminate the problems of having not having up to date electrical schematics But the major benefit is that these onboard schematics can be used by a computer program to support the bodybuilder as well as the workshop in their work with the additional equipment This includes services for verifying testing the logical design by simulation as well as introducing a service for troubleshooting the additional equipment The usage of model based diagnosis 2 7 to trou bleshoot technical systems is an active research area and techniqu
11. quipment to be configured to work in conjunction with the base platform Obviously interfaces are needed so that the equipment and base platform can work together the interfaces needed differ from application to application In this paper we focus on the base platform of heavy trucks and describe how we facilitate bodybuilder s equipment to work in conjunction with this platform Heavy trucks do not have a high demand for fast changes upon the added equipment in contrast to for example military planes which must be able to swiftly configure the platform to the payload and to the mission at hand But nevertheless is should still be possible to change and or adjust the equipment to cope with new customer demands or totally new demands from a new owner of the truck Usually this has been handled by providing the body builders with a dedicated Electronic Control Unit ECU which they could use to interact with the truck s Controlled Area Network CAN by pre defined control and informa tion signals We will in the reminder of the text use the term DE for the dedicated ECU provided to bodybuilders The bodybuilder then could define logical expressions typically using relays to control the usage of their equip ment By using relays and the DE the bodybuilder could create functions with desired behavior A consequence of having the logic outside the DE and often realized through electrical switches and relays made it difficult for worksh
12. rocess of adding their equipment to the truck as well as reducing the effort of designing the logical expressions and verifying them The benefits to a workshop are swifter and easier troubleshooting of not only the truck itself but also the bodybuilder s equipment A further improvement of the troubleshooting would be to use fault frequency statistics over components when ordering the sets The ordering could then be improved by present ing sets containing components with higher probability of faults before components with lower probability This could hopefully speed up the fault isolation process even more Yet another possible improvement would be to create dynamic troubleshooting guides that make use of the result from our troubleshooting and the set of Diagnostic Trouble Codes DTCs that are present on a particular truck REFERENCES 1 C E 50325 4 CSN EN 50325 4 Industrial commu nications subsystem based on ISO 11898 CAN for controller device interfaces Part 4 CANopen ISO Geneva Switzerland 2 J Biteus E Frisk and M Nyberg Condensed repre sentation of global diagnoses with minimal cardinality in local diagnoses in 17th International Workshop on Principles of Diagnosis DX 06 Spain 2006 R E Bryant Graph based algorithms for boolean function manipulation IEEE Trans Comput vol 35 no 8 pp 677 691 Aug 1986 ISSN 0018 9340 DOT 10 1109 TC 1986 1676819 Online Available ht
13. s Ws UPCONSSET ws end if end for until No more unique sets return Ws end function Le lt PARSETOINTERNALFORMAT logExpression Ws lt PARSETOINTERNALFORMAT wishedState C s PARSETOINTERNALFORMAT currentState Sets SAT Le Ws LabeledSets LABLESET Sets C s return SORT LabeledSets The inputs to the algorithm are the logical expression the wished state and the current state Output is a sorted list of sets containing ports for further investigation The function assConVal assigns values to the variable of the wished state that are consistent with the posted constraints The function upConsSet updates the wished state i e the variables in the constraint store The rest of the functions in the algorithm have self explanatory names First all three 3 input are parsed to internal format The SAT solver then finds all possible assignments set of sets given the wished state and the logical expression For each set the labeling function marks which ports that is inconsistent with the current set The assignments are then set of sets sorted with regard to marked ports The feedback the user receives is information on where the fault lies If the algorithm returns an empty set of sets then the additional equipment needs further investigation Otherwise the truck and or the additional equipment need further investigation The first set is presented to the user which since it is ordered has a minimum of inconsisten
14. s We have a truck with an EDE and logical expression s loaded on to it The bodybuilders additional equipment on the truck does not function as intended but it has previously worked fine The truck is in the workshop and we have our computer program connected to it and the truck is put in a state where the usage of the additional equipment should work We can then support the workshop in finding out whether A the additional equipment is faulty B the truck is faulty C both the truck and the additional equipment is faulty As mentioned before a consistency formulation of MBD is as follows Consistency formulation SD U COMPS U OBS f L Our problem setting only need to consider the input and output ports PORTS for potential abnormal behavior AB i e faults Hence we can change the problem formulation slightly by dividing the components into ports and operators OP c e COMPS c PORTS ce OP It is only the ports that can be either in abnormal mode or not p Ports AB p 8 AB p While operators in our setting are always error free o OP AB o The program in this setting do not have any knowledge about the world outside the logical expression When doing troubleshooting the program hands over the conclusions from the troubleshooting at the end of programs world i e the ports The program will assist the workshop in pointing out which ports conditions that are not fulfilled for a certain bodybui
15. set of sets where each set contains exactly one truth assignment to all the involved ports The SAT solver is hence used for the given logical expression and the values on the ports observation in form of J true 0 false or unbound value to calculate all possible values for the wished situation The output is a set of sets where each set is unique and consistent given the wished situation For each set in the set of sets from the wished state we calculate the difference or inconsistencies between the set and the current state We then label each set with the number of inconsistencies When this labeling is done we sort the set of sets in ascending order with regard to inconsistencies We can then present this information to the user in this order One can regard an inconsistent value as one or more possible faulty component s Using the common assumption that fewer faulty components or a simple hypothesis are more likely and as a consequence they are more likely to explain the real fault Which make sense if we regard that the nature of faults in our components are rare hence it is more probable that few components fail at the same time In algorithm 1 a more formal specification of the algo rithm is given Algorithm 1 The troubleshooting algorithm Inputs logExpression wishedState currentState Outputs SortedLabeledSets function SAT Le Ws repeat for all ws Ws do if ws then ws lt ASSCONVAL Le ws W
16. t ports faulty components in it Here an inconsistent item s is presented to the mechanic it can be physical pins on the EDE or CAN signals If it is a CAN signal for example that the signal parking brake applied is not present this can be further investigated by the mechanic The cause could for example be a faulty sensor If it is a physical pin on the EDE the mechanic can follow the attached wire and investigate the connected equipment If the mechanic cannot solve the problem using the information presented after checking each lead the next set is presented to the mechanic This process continues until the faulty component s are isolated IV PERFORMANCE To test the performance of the algorithm we set up an experiment where we used a block consisting of two 2 AND operators connected to one 1 OR operator see figure 4 ee BCI master C Pin 1 High ja BCI master C Pin 2 High PS pi BCI master B Pin 1 3 BCI master C Pin 3 High e rey ie BCI master C Pin 6 High Figure 4 One block These blocks are then connected to each other in layers e g where one 1 output block that has four 4 input blocks is considered as two 2 layers to form bigger and bigger logical expressions When testing the performance we ran out of unique inputs when using the above logical expression with three 3 layers as the prototype EDE could not handle more input We had to re use a majority of t
17. t be given so that the correct list of available signals is used Programmable signals and pins can be used for input and output and their colors define what type of signal they indeed are The input and output signals and their possible connections are shown in table I Table I HOW SIGNALS CAN BE COUPLED Input Output Digital or PWM Control signal Digital Pin 1 1 Information signal 1 1 PWM 1 0 In the table a 7 denotes that it is possible to connect the two signal types The current operators with the exception of the branch operator the rightmost symbol are only valid for digital Pins Information and Control signal types They symbols represent the usual interpretation of NOT OR AND and BRANCH operators loxi Vehicle speed above 5 kph Figure 2 The simulation view The software checks each user action involving the can vas so that only syntactically correct connections can be made and checks are made if the logical expression is complete or not Feedback about completeness is given to the user in form of circles colored either green or red The feedback is given on both the whole logical expression and on the used input output connections and for each operator used In the canvas of main window above we can see a small logical expression The expression uses two information signals from the vehicle to make sure that the vehicle speed is less than 5 km h
18. tp Idx doi org 10 1109 TC 1986 1676819 K Claessen N E n M Sheeran N S rensson A Voronov and K kesson Sat solving in practice with a tutorial example from supervisory control Dis crete Event Dynamic Systems vol 19 no 4 pp 495 524 2009 L Console and O Dressier Model based diagno sis in the real world lessons learned and challenges remaining in Proceedings of the 16th international joint conference on Artificial intelligence Volume 2 ser IJCAT 99 Stockholm Sweden Morgan Kaufmann Publishers Inc 1999 pp 1393 1400 Online Avail able http dl acm org citation cfm id 1624312 1624416 LabVIEW 2012 Labview system design software Online Available http www ni com labview R Reiter A theory of diagnosis from first principles Artificial Intelligence vol 32 no 1 pp 57 95 Apr 1987
Download Pdf Manuals
Related Search