Home

December 2012 - Minnesota Department of Transportation

Contents

1. 32 Figure 31 Weigh Pad plot utility ayle eya el sell 33 Fig re 32 Site Setup VW sa e halada Ayas ambyekaa imei ORS OE AAT ia aeta 34 Figure 33 Tools needed for WPad installation 0 al al RA R 36 Figure 34 Weigh padinstallationatCotton Minnesota TH 53 37 Figure 35 Sleeve anchor screws are fastened in 2 ft spacing sessssessesessssessessrssressessrssresseese 38 Figure 36 Some portions had wrinkles that caused vibration and error on the axle signal 39 Figure37 lLocationofweigh padaircavity 41 Figure 38 Installed weigh pads withaircavityandthetestvehicle 4 Figure 39 AircavitygeneratednoiseonChannel 0 C0 42 Figure 40 Removing Channel 0 signalclearsthesuperflucoussignals 42 Figure4l Chargeampsignalsofavan 0ct16 2010 44 Figure 42 Charge amp signals of a five axle semi trailer truck Aug 16 2011 44 Figure 43 Force and slope vela an Sip ksa elaman Sar iel izi dial dei 46 Figure 44 Speed effect test setup at MnRoad the weigh pads were fastened by high strength tapes on leadingandtrailingedgesandcarpettapesatthebottom 47 Figure 45 Scatter plot of speed vs weight of the same vehicle and linear regression
2. 16bit 200K S s SVGA 800 x ADC 600 LED back light LCD l LCD SW Type K USB ue Keyboard based Atom Based Mini ITX Mouse Thermocouple Master SW Li lon Polymer d Battery Bank erste es ACA 14 8V 9 Figure 19 Console computer block diagram 21 The enclosure design of the console box is shown is Figure 20 A local sheet metal fabricator constructed the enclosure using aluminum sheets 1 4 thickness which is shown in Figure 21 The inside of the box consists of three vertically stacked layers of compartments In the bottom layer two 14 8V Li Ion Polymer 16 000 mAh batteries and a USB thermocouple are enclosed along with an AC input plug for the battery charger Figure 22 shows the middle layer As shown in the picture a mini ITX board charge amps 2 5 SATA hard disk and a PCI ADC board are mounted Figure 23 shows the top layer Note that an SVGA LCD monitor a keyboard with a mouse pad a shut off master switch a shut off LCD monitor switch and a battery charger are mounted A reset switch is placed on the lid of the enclosure for an easy access It can be seen in Figure 24 as lit LEDs in the left top corner A USB hub is also available and is placed under the LCD monitor see Figure 23 The USB hub was added to allow easy download of the collected WIM data using a USB flash drive Figure 24 shows the back side of the enclosure There are four BNC connect
3. 48 Figure 46 Logregressionofweightdatabydifferentspeed5 49 Figure 47 Weigh pad installation at the northbound of TH 53 at the Cotton Minnesota 50 Figure 48 Scatter plot of IRD vs Weigh Pad GVW data 54 Figure49 ScatterplotofIRDvs Weigh Padspeeddata 54 Figure 50 ScatterplotofIRDvs Weigh Padvehiclelengthdata 55 Figure 51 Classification comparison between the IRD and Weigh Pad system vehicle records 58 List of Tables Table 1 RoadTrax BL Sensor Specific AS ia kasa enla ini saplama akay mahalli saklad n 6 Table 2 Weigh Pad Material Specifications i a li sessed cus onesies Soe la Vk 7 Fable 3 Console Mother Board ii sevse 80 0000000000240 A S ERES N e 20 Table 4 Key Console Box Components iile kei aka e ilm le 21 Table 5 Weigh Pad CSV Column Format eee da ak a lasa han gi Sa da 35 Table 6 Weigh Pad Installation and Removal Time 39 Table 7 Linear Calibration Factors multiplication factors for Different Speeds 48 Table Toe Calibration FaCters sa das amiyane kadim kre Kakma asi sapa sl aka amel sisi 49 Table9 Sep Parameters us alm ae ee anti Sal roves BM ae 51 Fable 10 limit Parameters orir cinnin aria G AR S YARAM el 51 Table 11 Correlation Coefficients and R Between IRD and Weigh Paddata 53 Table 12 Av
4. 2 45 2 18 4 63 2 0 80 44 3 30 1 11 41 07 0 66 4455549799977997722110 00 15 9 0 46 2 31 1 11 41 27 2 67 8 9 2 61 2 63 5 24 2 0 58 47 9 C 3 32 2 11 41 34 3 70 10 5 12 5 1 98 1 67 1 32 4 97 3 0 63 48 3 33 1 11 42 36 0 66 4555419722900997221170 00 15 8 0 44 0 34 1 11 42 37 2 67 9 4 2 49 2 16 4 65 2 0 43 43 8 35 1 11 42 47 2 69 9 4 2 58 2 04 4 62 2 0 75 43 2 36 1 11 42 55 5 65 12 1 4 3 28 5 4 1 10 71 14 60 13 52 13 24 11 68 63 75 9 0 44 42 7 37 1 11 43 17 2 66 19 7 9 40 10 45 19 85 5 0 92 41 7 38 1 11 43 29 2 48 14 6 17 15 12 22 29 37 5 0 66 41 4 39 1 11 43 43 2 71 11 8 2 84 2 17 5 01 3 0 8 41 1 40 1 11 44 47 5 65 18 4 4 3 28 6 4 1 3 76 3 55 3 26 3 00 2 97 16 54 9 0 72 45 9 41 2 11 44 47 5 65 18 4 4 3 28 8 4 1 4 72 3 69 3 43 2 83 2 88 17 54 9 0 71 45 9 43 1 11 45 18 2 67 13 7 5 82 11 41 17 23 5 0 43 46 7 44 1 11 46 37 2 71 9 4 4 33 2 70 7 03 2 0 54 45 6 45 1 11 46 40 4 64 10 8 20 8 2 6 2 96 3 34 2 66 3 13 12 09 3 0 16 45 3 46 2 11 46 48 2 69 11 7 3 38 2 54 5 92 3 0 58 44 7 47 1 11 47 03 2 71 8 7 2 48 2 56 5 04 2 0 63 45 1 48 1 11 47 18 2 65 9 3 2 31 2 55 4 86 2 0 47 46 8 49 1 11 47 42 2 61 9 1 3 12 2 82 5 94 2 0 92 49 3 50 2
5. A block diagram of the computer developed for the weigh pad console is shown in Figure 19 The system s computing is powered by an Atom based Mini ITX board specified in Table 3 The ADC is interfaced through a PCI bus and an SVGA LCD monitor is connected through a VGA port The SVGA monitor has a resolution of 800 x 600 pixels which is low in today s graphic standards but is sufficient to provide a rich graphical user interface GUI for the current portable WIM system The LCD monitor is an open frame and LED back lighted LCD designed for outdoor applications The screen was indeed easily readable under sunlight Once the system is initialized and if it is in a run state that no longer requires visual human interface the LCD monitor is turned off by a hard on off switch on the power supply line A thermocouple is interfaced through a USB port of the motherboard and is used for measuring pavement temperature A summary of key console components is summarized in Table 4 It should be noted that all of them are off the shelf products Table 4 Key Console Box Components Component Product Model Manufacturer Motherboard Custom M350 Mini ITX Logic Supply ADC PCI DAS6013 Measurement Computing USB Thermocouple USB 2001 TC Measurement Computing Type K thermocouple SC GG K 30 36 Omega LCD monitor LBT 10420 1 4 Caltron Industries Battery Charger TLP 2000 Tenergy Battery 14 8V Li Ion Tenergy Polymer 16Ah
6. 35 Chapter 4 o Weigh Pad Pavement Installation One of the challenges of developing a practical portable WIM system is to design a simple and effective installation procedure that securely attaches the weigh pads on the pavement After installation the sensor pads should retain the tightness to avoid vibration endure tire traction forces and provide accurate readings of the axle loads for the duration of the data collection Because the sensors must be quickly installed and removed in portable applications it is a challenge to develop a good installation method Any movement of sensor position or vibrations would decrease the accuracy of the sensor readings This chapter describes how to install weigh pads on highways and discusses issues related to installation 4 1 Installation on Highways Initial installation and driving tests were conducted using only one or two test vehicles from the low volume road at the MnRoad facility The installation methods tested are 1 Gorilla tapes a strong bonding utility tape on leading and trailing edges of the weigh pad 2 carpet tapes at the bottom and Gorilla tapes on leading and trailing edges of the weigh pad 3 carpet tapes at the bottom Gorilla tapes on leading and trailing edges and concrete screws on the center of the weigh pad 4 Gorilla tapes on leading and trailing edges and sleeve anchor screws with a flat washer in the middle of the weigh pad All four approaches were tested at th
7. Minnesota Department of Transportation Development of a Weigh Pad Based Portable Weigh In Motion System RESEARCH SERVICES Office of Policy Analysis Research amp Innovation Taek M Kwon Principal Investigator Department of Electrical and Computer Engineering Northland Advanced Transportation Systems Research Laboratories University of Minnesota Duluth December 2012 Research Project Final Report 2012 38 To reguest this document in an alternative format please contact the Affirmative Action Office at 651 366 4723 or 1 800 657 3774 Greater Minnesota 711 or 1 800 627 3529 Minnesota Relay You may also send an e mail to ADArequest dot state mn us Please request at least one week in advance Technical Report Documentation Page 1 Report No 2 3 Recipients Accession No MN mn 4 Title and Subtitle 5 Report Date Development of a Weigh Pad Based Portable Weigh In Motion C 7 Author s 8 Performing Organization Report No Taek M Kwon 9 Performing Organization Name and Address 10 Project Task Work Unit No Department of Electrical and Computer Engineering CTS Project 2009020 University of Minnesota Duluth 11 Contract C or Grant G No 1023 University Drive Duluth MN 55812 C 89261 WO 114 12 Sponsoring Organization Name and Address 13 Type of Report and Period Covered Minnesota Department of Transportation Final Report Research Services 14 Sponsoring Agency Code 395 John Ireland Bou
8. For this comparison the weigh pads were installed on the northbound two lanes using the installation method described in Section 4 1 The office of Transportation Data amp Analysis TDA at MnDOT requested the lane closure and supplied the tools needed for installation The lane closure was provided by District 1 personnel from their maintenance office Installation of weigh pads for the two northbound lanes took 30 minutes Figure 47 shows a picture of the TH 53 weigh pad installation and a vehicle passing through in lane 2 It is not visible in this picture but the Kistler sensors are located about 18 ft 5 49 m before the weigh pads Figure 47 Weigh pad installation at the northbound of TH 53 at the Cotton Minnesota The spacing between the upstream and downstream weigh pads was 14 ft 4 267 m sensitivity was set to the default value of 6 25 and the weight data was not calibrated i e Calibration Factor 1 for all sensor segments Table 9 summarizes the setup parameters The limit parameters required by the weigh pad console are summarized in Table 10 The data was collected between 11 3 2011 11 21 57 AM Thursday and 11 4 2011 9 55 27 AM Friday The road surface was covered with frosts in the early morning but cleared at the time of installation cleared around 8 30am 50 Since the sensor pads stayed overnight a good portion of time the system was in a frosty condition which was a concem for physical sensor pad damages but
9. The weigh pad system produces text based CSV files for the WIM data The filename follows a format that consists of the date and site ID text strings as follows yyyymmdd csv where yyyymmdd is year month day and is the three digit site ID The column format is summarized in Table 5 The columns up to 32 are identical to the BullConverter column format The column 33 which is the pavement temperature is only available for the weigh pad system The CSV file is stored in the yyyymmdd folder of the data root path defined in the Site Setup menu 34 Table 5 Weigh Pad CSV Column Format Column Column name Description Number l Index Vehicle record is numerically indexed and this column shows the index 2 Lane Lane number of the vehicle passed through 3 Time Time in hh mm ss where hh is military hour 4 AxleC Number of axles on the vehicle 9 Speed Speed of the vehicle in mph 6 16 Axle Spacing AS Axle spacing in feet It contains total 11 fields ASI AS11 separated by comma 17 28 Axle weights AW Axle weights in Kips It contains total 12 fields AWI AW12 separated by comma 29 GVW Gross Vehicle Weight GVW in Kips It is simply a summation of each axle weight 30 Class Vehicle class determined by the classification algorithm 31 Err It is a numeric code that represent an error 32 100thSec 100 seconds of the time in Column 3 33 pavTemp Pavement temperature in Fahrenheit
10. Weigh in motion No restrictions Document available from scales Weighing devices National Technical Information Services Alexandria Virginia 22312 19 Security Class this report 20 Security Class this page 21 No of Pages 22 Price Unclassified Unclassified 87 Development of a Weigh Pad Based Portable Weigh In Motion System Final Report Prepared by Taek M Kwon Department of Electrical and Computer Engineering Northland Advanced Transportation Systems Research Laboratories University of Minnesota Duluth December 2012 Published by Minnesota Department of Transportation Research Services 395 John Ireland Boulevard MS 330 St Paul Minnesota 55155 This report documents the results of research conducted by the authors and does not necessarily represent the views or policies of the Minnesota Department of Transportation or the University of Minnesota This report does not contain a standard or specified technique The authors the Minnesota Department of Transportation and the University of Minnesota do not endorse products or manufacturers Trade or manufacturers names appear herein solely because they are considered essential to this report Acknowledgements This research was supported by the Minnesota Department of Transportation MnDOT The author would like to thank the technical liaison Ben Timerson and the Technical Advisory Panel Mark Novak Josh Kuhn and Bruce Moir for providing suggestions a
11. and the corresponding axle waveforms along with vehicle records computed by the weigh pad software were recorded The test vehicle was driven over the weigh pads a total of 56 times The data was collected from both Lane 1 and Lane 2 However the weight data from Lane 2 was noisy due to unstable pad installation so Lane 1 data was used for the analysis 46 Figure 44 Speed effect test setup at MnRoad the weigh pads were fastened by high strength tapes on leading and trailing edges and carpet tapes at the bottom 5 2 3 Analysis As the first step of the analysis an x y scatter graph of speed vs weight was plotted to observe the data trend Figure 45 shows the data points of speed vs weight along with a linear regression line The data clearly shows an increasing trend of weights as a function of the speed For example the average weight is 5 Kips 2 27 ton at 30 mph but the average weight becomes 6 5 Kips 2 9 ton at 60 mph Linear regression of this data was computed and shown along with the x y scatter graph and the equation is given by y 0 0533x 3 2478 Kips 11 where x is the speed in mph and y is the vehicle weight in Kips This function has R 0 7897 R is a measure of goodness of fit 0 representing the worst fit and 1 representing the best fit Calibration or multiplication factors for different speeds could be obtained from this regression which is summarized in Table 7 The relationship between the calibrated
12. traffic engineers need to know the truck volumes and weights but without the cost of permanent roadside WIM stations One solution to bring WIM technology to local roads is to utilize a portable WIM system much like pneumatic tube counters used in short duration traffic counts That is a single unit is reused in multiple locations for few days at a time This way WIM data is obtained without the cost of a permanent WIM station Unfortunately WIM development efforts have mainly been given to in pavement permanent systems consequently portable WIM systems are not available on the market This report describes the results of a two year research project sponsored by the Minnesota Department of Transportation MnDOT to develop a portable WIM system that can be readily deployed on local roads The objective of this project was to develop a portable WIM system that would be used much like a pneumatic tube counter The sensor chosen was the RoadTrax BL sensor strip or simply BL sensor which is a thin narrow piezoelectric strip To accomplish the project objective the BL sensor strips were sandwiched and glued between two strong conveyer belts Conveyer belts provide flexibility and durability needed for on pavement installations A standard sensor constructed has a length of 24 ft covering two lanes of roads and a width of 1 ft This new sensor is called a weigh pad The final completed system is called a weigh pad system and consists o
13. 7 percent would affect vehicle classification Figure 51 shows the vehicle classification results of the two systems using blue and red bars No class 1 vehicles were detected during the test period so the classes shown are from class 2 to class 13 Total 49 vehicle records out of 3 235 records were classified as different classes between the IRD and weigh pad WIM systems This difference in classification is only 1 5 Also the differences of classification were mostly within the neighboring classes such between as class 2 and class 3 Since the axle spacing difference is minimal as shown in the previous data 1 5 classification difference must have caused by the weight differences It should be mentioned that the weigh pad data was not calibrated Therefore only 1 5 difference should be considered that classification results by the two systems are remarkably close 57 1600 Classification Comparison 1400 1200 mIRD 1000 m WPad Cc 5 O 800 o i gt 600 j i mo ual 12 i a 1 33172 3 3 1 2 Figure 51 Classification comparison between the IRD and Weigh Pad system vehicle records 58 Chapter 6 Conclusions and Future Recommendations 6 1 Conclusions This report presented the results of a MnDOT sponsored project on the development and evaluation of a portable WIM system referred to as a weigh pad system The original objective was to bring WIM technologies to rural local roads by developing a low
14. The second part of the circuit is a DC servo loop that consists of U2 C3 R5 C2 and R6 This circuit pulls the DC level down close to the signal ground The resistor capacitor value relationship in this circuit should be C2 R6 C3 R3 and the passive component values must be determined based on the time constant required Since the time constant of the DC level removal was set at 20 seconds component values RS R6 10M Ohm and C2 C3 2 2 uF were selected as the final values Two channel charge amp circuits were constructed according to the Figure 13 design Figure 14 shows the final PCB with components soldered onto the board For the PCB design the Mentor Graphics PADS software tool was used The PCB was manufactured from a PCB prototype outlet 16 Vs R6 R4 7 U2 R2 R3 Vs R5 C3 p C1 Vs R1 Vout U1 O WIM T Sensor Vs Figure 14 A prototype two channel charge amp built for this project 2 4 Analog to Digital Converter ADC What type of ADC to be used is one of the key decisions that must be made based on the choice of the computing system console computer required sampling rate and signal resolution The ADC choice often falls into two types 1 USB based ADC or 2 PCI based ADC Initially the PI considered a laptop computer as the console computer of the weigh pad system in which case USB based ADCs are the natural ch
15. cost WIM system that would be portable and reusable much like pneumatic tube counters After many trials and errors the final working sensor called weigh pad was constructed by gluing piezoelectric sensor strips between two thin and long conveyer belts A standard weigh pad has a length of 24 ft 7 3 m covering two lanes and a width of 1 ft 0 3 m For installation two weigh pads are laid across the traffic lane separated by a known distance typically 12 16 ft or 3 7 4 9 m and fastened on the pavement surface using sleeve anchor screws The edges are then taped using strong bonding utility tapes Installation takes about 15 minutes per lane while removal takes about 7 minutes per lane The developed weigh pad system was tested in a number of different ways Among them a side by side test with an in pavement permanent station on a truck highway provides a meaningful comparison According to the data comparisons axle spacing and speed were nearly identical 0 5 different while GVW was about 4 different in NRMSE measurements Comparison of vehicle type classification revealed a difference of only 1 5 All of the comparative numbers presented in Chapter 5 suggest that data quality of the weigh pad system is within a few percentage points of in pavement permanent systems This project successfully demonstrated that a reusable portable WIM system that would be installed much like a pneumatic tube counter can be built A side by side comp
16. extremely easy The drivers were stable and facilitated the claimed sampling rate If a USB interface was to be used for the weigh pad ADC the USB 6210 board would have been the best choice among the three tested Figure 15 Measurement Computing USB 7202 200K S s 16 bit ADC Figure 16 Access I O Products Inc USB AI16 16A 500K S s 16 bit ADC 18 Figure 17 NI USB 6210 200K S s 16 bit ADC In the end a laptop computer platform was not selected as the console computer because of their energy consumption which is discussed in the subsection 2 5 The final console computer selected for this project was a single board computer based on an Intel dual core Atom processor Such boards consume much less energy than common laptop computers and come with a PCI interface slot The PCI interface provides a higher data rate transfer than USB 2 but more importantly it provides a reliable data transfer and proven software drivers that have been used in the field for a long time On the other hand USB boards have high overheads and the drivers are often unstable when the ADC continuously runs with a high sampling rate for long hours more than 24 hours The final PCI board selected for the weigh pad ADC was PCI DAS6013 manufactured by the Measurement Computing Inc and its picture is shown in Figure 18 This board provides a true 200kS s at 16 bit resolution up to 16 analog inputs channels The same type of boards has been used by the PI i
17. if the signal is continuous the signal area and weight are related as sig W souna x f pound t Thus W poundtw 6 sig f pound Equating Eqs 5 and 6 gives W souna t aCN IT pound Finally a is computed as W Sunal pound w 7 dam 7 Collecting these results the final axle waveform with computed parameters of a and cis obtained as a 27 2 fQ ae th for 0 lt x lt 1 where a and c are give as W onat pound w a f pound az l w c 5c Example 1 This example shows how the actual numerical values from the given model are generated for the DAC Assume that the following axle characteristics are given Axle Weight Wna 20 000 pounds Sampling Rate R 4000 S sec Footprint Length 0 656167979 feet Speed v 88 foot sec Sensor width w 0 16404199475 feet Multiplication factor f 1541 4768 2 27 Using Eqs 7 and 3 the parameters a and c are calculated as a 3 66 c 0 003728 For sampling rate 4000 S sec the digitized signals of f x are generated as For i 0 to 3999 x i 2000 4000 shift of signal to right 2000 points fa sac The plot of these values is shown in Figure 27 It should be noted that the non zero data image is located at 2000 2 5c 4000 lt i lt 2000 2 5c 4000 Voltage 1 800 1 900 2 000 2 100 2 200 Digital Sequence Number Figure 27 Plot of the Example 1 axle signal 3 1 2 Computing th
18. leading and trailing edges of the pads are tapered by sanding the edges to create a smooth slope The measured thickness of the weigh pads when two 908860 pads are sandwiched together was 0 295 in 7 5 mm This would be the thickest part of the weigh pad The total weight of a finished single lane single strip weigh pad was 11 4 Ibs 5 17 Kg Figure 3 shows a single lane single strip weigh pad constructed according to the specification in Figure 2 The weigh pad can be easily wrapped around as a loop as shown in Figure 2 for easy carrying The length of the coaxial cable is 100 ft Figure 2 Dimensions of single lane single strip weigh pad Figure 3 A single lane single strip weigh pad prototype In order to be a complete WIM sensor a pair of sensors is needed More specifically two identical sensor pads must be installed separated by a known distance to compute the speed of the weighing vehicles The vehicle speed is used to normalize the axle waveforms so that the weight of a vehicle is independent of its speed If two sensors are needed one way of making the installation convenient is to embed two sensor strips in parallel in a large single pad This idea was tested and the design of embedding two sensor strips in parallel is shown in Figure 4 The actual prototype constructed for the Figure 4 specifications is shown in Figure 5 We refer this sensor as a single lane dual strip weigh pad The signals are carried by t
19. of the sensor guality Weight measurements of moving vehicles are often most inconsistent because vehicles have a suspension system that oscillates the weight over a time and space In order for two weight measurements to be identical the tension of the vehicle suspension must be identical as well as the speed and wind condition The weigh pads and IRD sensors were in a close proximity 18 feet apart but the tension of individual vehicle suspension at two different positions is a factor that cannot be controlled An RMSE or NRMSE measurement provides a single numerical representation of average one to one differences thus they do not show the details of data trends or relations within In order to investigate the trends or relations in data scatter graphs between the IRD vs weigh pad data on GVW speed and vehicle length is plotted along with computation of the correlation coefficients see Figures 48 50 The correlation coefficient of X and Y denoted by Corr X Y is defined by Cov X Y o0 Corr X Y 16 where Cov X Y is the covariance between two random variables X and Y and o and o are the standard deviations of X and Y If Corr is closer to 1 two random variables are more strongly linearly correlated In general Corr gt 0 8 suggests a strong linear relationship 20 The coefficient of determination denoted as R is also computed 20 This value represents a statistical measure of how well the regression line approxima
20. passes through a weigh pad the weigh pad acts as a small bump that creates a vibration This vibration is propagated to CO but it is amplified by the air cavity before it reaches to the CO sensor When an axle hits the air cavity the air is compressed and then it is transferred to the sensor strip Another important factor is that piezoelectric materials generate high amplified signals if vibration is within the range of its resonant frequency It appears that CO signals in Figure 39 were in the range of sensor resonant frequency since the amplitude of the signals is as high as real loads Clearly the air cavity problem is avoidable if the pads are carefully manufactured without any air cavities Indeed the second set of weigh pads was produced with caution and did not have any superfluous signal problems when it was tested on the same road using the same test vehicle This new sensors were used in the final tests 40 bit A x Weigh Pad 1 Ii Air Cavity T I 11 ae pi iol take RoadTrax id 143 BL Sensors 13 9 pig E Test Vehicle l Travel Direction ji Ch2 Y v cht TT Weigh Pad 2 ri Yellow Lane Markings Figure 37 Location of weigh pad air cavity Figure 38 Installed weigh pads with air cavity and the test vehicle 41 Volt 0 6 14 000 14 500 15 000 15 500 16 000 16 500 17 000 17 500 18 000 Sample Index Figure 39 Air cav
21. 0 5914 The weigh pad system computes the ESAL of each vehicle using the parameters set by the ESAL Setup window This window appears when the menu item ESAL Setup is selected The default values are shown below a ESAL Setup lo x Pavement Type Flexible Rigid Flexible Structure Number SN E Terminal Serviceability Pt s 0Ctti i Rigid N S Bo Save Changes Slab Thickness in Inches D Terminal Serviceability Pt 25 Exit The changes become effective only after the parameters are saved and then activated or when the Bulldog is restarted The axle detection is done when the charge amp signal is greater than the threshold value added to the signal idle resting level In the real implementation the beginning of the axle start signal is traced back from the threshold detection Because the signal condition of each channel can be different a threshold value is set for each channel This window is selected from the menu item Signal Thresholds B 2 a Axle Signal Threshold Setup E ol x Lane 1 Upstream ChO 0 12 Volt Lane 1 Downstream Chi 0 12 Yol Lane 2 Upstream Ch2 foi2 Volt Lane 2 Downstream Ch3 0 12 Volt lt a ea The weigh pad software utilizes several limit parameters supplied by the user For example the parameter Maximum axle spacing possible is used to determine the boundary between vehicles The limit parameters are supplied through the Parameter Limit
22. 000 at the time of this writing In addition maintenance of a WIM site requires a recurring cost of sensor maintenance electricity communication and system upgrades Nevertheless building WIM stations could be readily justified for heavy traffic highways such as interstate and trunk highways Unfortunately installing a WIM station on a rural local road is rarely justifiable considering that rural roads have limited Average Daily Traffic ADT Low ADT does not mean fewer overweight violations or diminish the need for protecting the roads from overweight vehicles In fact heavy truck traffic volumes on local roads of Midwestern states have been increased due to new demands on renewable energy or ethanol related agricultural products such as corn and soybeans This increased truck volume raises a grave concern for many local transportation engineers as it could significantly shorten the life of the local roadways To estimate the road wears or to protect from overweight vehicles it is essential to collect truck weight data Therefore local roads with load ADT also need WIM stations One solution to bringing WIM technologies to local roads is to develop a low cost portable WIM system that can be used as a short duration WIM data collection system similarly to the use of pneumatic tube counters Portable WIM systems could be used for just few days in the area where frequent weight violations likely occur There are several benefits of using a portabl
23. 10 MnRoad is a test track owned and operated by MnDOT for evaluation of new pavements or road sensor technologies The setup and test pictures are shown in Appendix A As shown in the picture a pair of single strip weigh pads and a dual strip weigh pad were installed side by side A 2005 Toyota van and a five axle semi trailer truck were used as the test vehicles Figure 6 shows a waveform of the Toyota van driven over the single strip weigh pad The data was sampled at 4 096 samples per second It clearly shows waveforms of the two axle load signals with some ripples in the beginning and at the end of the axle signals The idle level of the signal between axles stays close to ground which is desired and important for threshold detection of axle load signals Before MnRoad tests several tests were conducted at the UMD parking lots All test results showed that single strip weigh pads are good enough to obtain stable axle load signals 10 3 2 8 2 6 2 4 2 2 2 1 8 1 6 1 4 1 2 1 0 8 0 6 0 4 0 2 0 0 2 0 4 0 6 35 000 35 200 35 400 35 600 35 800 36 000 36 200 36 400 36 600 36 800 37 000 Sample Index v c0 Cl Volt Figure 6 A waveform generated by a single lane single strip weigh pad for a Toyota van Waveforms of a semi trailer truck a test truck available at MnRoad generated from a single lane dual strip weigh pad are shown in Figure 7 In the graph the channel 0 Ch 0 signals come from the leading u
24. 11 47 44 2 67 11 7 4 03 4 13 8 16 3 0 19 49 5 51 1 11 48 07 3 57 18 4 4 5 11 06 15 75 14 48 41 30 6 0 91 51 0 52 1 11 48 20 3 60 12 0 13 2 3 57 4 08 2 34 9 99 3 0 57 50 3 53 1 11 48 22 0 60 54 54473712317777p 0 00 15 8 0 50 3 C 4
25. 2 61 0 54 5 60 3 0 50 49 5 8 1 11 35 28 2 65 10 2 3 35 2 81 6 17 3 0 71 48 6 9 2 11 35 29 2 68 12 2 2 99 1 71 4 70 3 0 5 48 6 10 1 11 35 30 0 67 55 19797979990017770 00 15 8 0 48 5 11 2 11 35 30 2 65 8 B 2 44 1 89 4 33 2 0 99 48 5 12 1 11 35 34 2 71 8 5 4 24 4 10 8 34 2 0 18 48 2 13 1 11 35 39 2 67 10 5 2 76 2 77 5 53 3 0 15 47 8 14 1 11 35 46 2 65 8 6 2 09 1 81 3 90 2 0 93 47 3 15 1 11 35 50 3 62 11 6 15 9 4 63 3 61 3 88 12 12 3 0 21 46 8 16 1 11 36 06 2 71 9 5 4 11 3 82 7 93 2 0 4 45 6 17 1 11 36 51 2 70 10 2 3 05 2 25 5 30 3 0 0 44 6 18 1 11 37 15 0 69 44534y37y7771 0 00 15 9 0 44 0 19 1 11 37 16 2 72 9 3 2 34 1 29 3 63 2 0 91 43 9 20 1 11 37 29 2 71 12 0 3 75 2 60 6 35 3 0 17 43 5 21 1 11 38 01 0 70 4455 99790900097722110 00 15 8 0 42 9 22 2 11 38 14 2 72 13 4 2 82 1 50 4 32 3 0 7 43 1 23 1 11 39 04 2 71 9 4 3 71 3 30 7 01 2 0 4 41 7 24 2 11 39 11 0 68 45555979990007722110 00 15 8 0 41 6 25 2 11 39 12 2 73 10 8 2 35 2 92 5 27 3 0 43 41 5 26 1 11 39 23 0 71 5455909009900007722110 00 15 8 0 41 3 27 1 11 39 31 2 65 11 2 2 41 2 31 4 71 3 0 91 41 3 28 1 11 40 19 0 68 4555 29799900000722110 00 15 8 0 44 2 29 1 11 40 39 2 64 9 6
26. 4 Weigh pad Veh Length feet D O 0 10 20 30 40 50 60 70 80 IRD Veh Length feet Figure 50 Scatter plot of IRD vs Weigh Pad vehicle length data In Figure 48 a scatter graph of IRD vs weigh pad GVW the data points are not tightly bunched to the linear line Initial assessment of this spread effect is that the vehicle weight is influenced by the springing effect of vehicle suspension This spring effect is independent of the sensor accuracy and should lead to random differences of weights between the IRD and weigh pad measurements within a certain range of GVW In order to investigate whether speed was a part of the cause of the spread in Figure 48 or not a two dimensional GVW ratio table with respect to speed was created and shown in Table 12 GVW ratio is defined by GVW Ratio GVW wpag GVW Rp 17 Here GVW wpag is the raw weigh pad GYW data that was not calibrated In the table speeds were spaced by 10 mph 16 Km h while the GVW was spaced by 10 Kips 4 535 Kg Each table entry is the average of the GVW Ratio values in Eq 17 in the defined range The number of vehicle records in each GVW Ratio bin for average computation is shown in Table 13 Because this data was collected from a Trunk Highway most vehicles are clustered around the speed range of 60 90 mph 96 6 144 8 Km h The average of GVW Ratio is 1 26 which indicates that weigh pad GVW is 26 higher However this trend is inconsistent and some cases it
27. 4 262 07 45 30 Upstream Loop Failure 546 Figure 28 Net component vehShow dll developed for visual modeling of individual vehicle records In addition there should be a table or a spread sheet of vehicle records so that the user can trace back a list of vehicle records This table should be equivalent to the actual vehicle records stored in the WIM data file In addition for a diagnostic purpose it is useful to have a real time plot of the charge amp waveforms This real time plot could be used like an oscilloscope to check whether the charge amp or ADC is working properly or not The plots could also be used for checking the resting voltage levels of the charge amp or line noise conditions At the end the following items were selected as the console functions A screen capture of the final GUI of the system is shown in Figure 29 Table of WIM vehicle records Real time display of pavement temperature A real time plot routine for the ADC raw signal A recording utility for the real time ADC binary data A text reading tool for the WIM vehicle record data GUI interfaces for setting of all sorts of parameters 30 W Bulldog WeighPad V1 2 Site 031 Two Harbors HW61 mp 16 4 x Fie Settings Graph Help 1707008 1 8 1 3 Kips Speed Class GVW ESAL Time Error 8 2 Feet S6mph 2 35Kips 0000 11 41 19 110 193 8 5 77 10 0 Kips Speed Class GVW ESAL Time Error 103 62 15a 5 1 Feet 7 mph 9 45 6Kips 0 237 11 41 19 Veh Time L
28. 6 20 80 92 02 9 0 66 35 7 15 1 11 25 42 2 65 11 5 7 15 5 57 12 72 5 0 30 35 5 16 1 11 25 53 2 67 10 0 3 88 3 44 7 32 3 0 67 35 6 17 1 11 25 54 0 61 55509079799000007770 00 15 12 0 35 6 18 2 11 26 06 0 20 55579990797772707770 00 15 8 0 35 5 19 2 11 26 07 2 72 9 5 5597112 772 97 yy 5497792719 19 2 0 80 35 6 20 1 11 26 20 6 67 15 0 4 5 24 5 4 8 5 0 12 98 10 48 9 58 8 49 9 59 10 75 61 87 10 0 73 35 4 21 1 11 26 29 6 67 16 7 4 3 18 7 4 8 4 8 13 16 18 63 18 60 12 58 22 28 20 49 105 74 10 0 61 35 4 22 1 11 26 37 2 65 10 1 5 15 5 45 10 60 3 0 55 35 3 23 1 11 26 44 2 66 9 0 3 15 2 28 5 43 2 0 24 35 3 24 1 11 27 05 2 69 11 7 4 73 4 03 8 76 3 0 89 35 3 C 1 25 2 11 27 15 2 71 9 7 1 86 2 14 3 99 2 0 10 35 3 26 1 11 27 17 2 68 9 3 1 63 2 77 4 41 2 0 36 35 3 27 1 11 27 34 2 71 8 7 3 82 3 18 7 00 2 0 42 35 1 28 1 11 27 43 2 62 11 4 3 22 2 89 6 12 3 0 83 35 2 29 1 11 27 45 2 69 8 8 2 17 1 55 3 72 2 0 38 35 2 30 1 11 27 53 4 65 21 2 23 5 2 7 9 60 29 94 8 39 8 56 56 49 4 0 36 35 2 31 1 11 28 12 2 69 9 9 4 51 3 84 8 34 3 0 7 35 3 32 2 11 28 21 0 46 5545590970990001072210 00 15 8 0 35 5 33 2 11 28 23 2 30 13 2 3 61 3 14 6 75 3 0 80 35 5 34 2 11 28 28 2 70 12 1 1
29. 600 pixels Although this LCD monitor s resolution is limited in comparison to today s high resolution monitors it is good enough to create an easy to use Graphical User Interface GUT for operation of the weigh pad WIM system For the software design of the weigh pad WIM system operational needs of the system and a list of required components were created first Visual modeling of vehicle records was one of the 29 requirements since a visual form can serve as an excellent verification or diagnostic tool for maintenance operations at the site i e the vehicle model in the screen can be visually compared with the actual vehicle The information displayed on the visual vehicle model includes axle weights axle spacing speed classification GVW ESAL time error message lane number lane direction vehicle identification number The method of visual modeling adopted in this research is developing a dll net component so that it can be drag and drop into any window The c language of the Microsoft Visual Studio provides an excellent tool for developing visual components and was used in this project The component named vehShow dll was developed and its visual interface is shown in Figure 28 In this vehShow component vehicle information items mentioned above are implemented as properties ee _ee penis w 20 0 18 0 200 200 15 0 Kips Speed Class GVW ESAL Time Error Veh 5 0 19 0 50 12 0 Feet 7Omph 9 93 0Kips
30. 72 0 68 2 40 3 0 71 35 6 35 1 11 28 30 2 65 11 6 3 04 2 52 5 56 3 0 21 35 6 36 1 11 28 46 2 67 9 4 2 87 2 15 5 02 2 0 87 36 1 37 1 11 29 29 6 64 16 2 4 3 18 6 4 9 4 9 14 30 17 89 18 36 9 42 19 82 20 27 100 06 10 0 13 45 5 38 2 11 29 32 2 69 8 6 2 30 1 97 4 26 2 0 0 47 0 39 1 11 29 55 3 70 11 6 15 4 3 33 5 07 2 16 10 57 3 0 26 48 6 40 1 11 30 16 5 69 17 8 4 3 29 8 4 0 16 06 20 56 22 73 23 02 6 89 89 26 9 0 16 47 5 41 1 11 30 28 3 68 10 5 15 5 1 97 2 76 1 65 6 38 3 0 33 46 9 42 1 11 31 13 2 67 9 8 5 31 5 59 10 90 3 0 15 45 4 43 2 11 31 13 2 69 8 7 1 44 1 27 2 70 2 0 56 45 4 44 1 11 31 17 0 62 5559000700977000020 00 15 12 0 45 4 45 1 11 31 38 2 64 10 2 3 68 2 58 6 25 3 0 42 45 4 46 1 11 32 12 2 69 12 0 3 29 3 40 6 69 3 0 90 45 6 47 1 11 32 20 2 70 11 8 3 83 2 89 6 71 3 0 12 45 9 1 1 11 33 10 2 72 9 4 3 56 2 48 6 04 2 0 29 46 8 2 1 11 33 35 2 77 10 1 3 04 2 64 5 68 3 0 61 48 3 3 1 11 33 42 2 70 10 6 1 72 1 62 3 34 3 0 23 48 7 4 1 11 34 24 0 66 4 551709970972000720 00 15 8 0 50 7 C 2 5 1 11 34 25 2 71 9 0 3 91 2 36 6 28 2 0 58 50 6 6 2 11 35 01 2 61 9 6 2 48 2 09 4 57 2 0 57 48 9 7 2 11 35 15 3 69 10 9 13 2 2 45
31. 860 Product Construction 2 Ply Filament Polyester Carcass Black smooth Rubber Cover Both Sides Color Black Compound Formulation Grade II Rubber Nominal Overall Gage inches mm 0 154 0 015 3 9 0 4 Nominal Weight in Ibs ft7 K g m 0 98 10 Rated Working Tension 160 Ibs in 28 N mm 2 Top Cover Surface Semi Smooth Bottom Cover Surface Semi Smooth Minimum Pulley Diameter 4 inches 102 mm Temperature Range 20 F to 225 F 29 C to 107 C Special Standards RMA Grade II Covers Cover Coefficient of Friction Steel 0 75 Nominal Bottom Coefficient of Friction Steel 0 75 Nominal Production Width 72 inches 1829 mm Manufacturer Forbo Movement Systems Product family Transtex 2 2 2 Single Lane Weigh Pads Design The first sensor built and tested was a single lane single strip weigh pad shown in Figure 2 First the 908860 belt was cut to two 6 x 144 in 15 x 366 cm pads A 144 in 366 cm long groove with the cross section of the groove size width x thickness 0 275 x 0 075 in 6 985 x 1 905 mm is made at the bottom pad conveyer belt The sensor strip is inserted to the slot and glued The top pad is next glued to the bottom pad During this process it is important to eliminate any air pockets between the glued pads and sensor strips because these air pockets can pop by a load and influence the BL sensor strips to create superfluous signals As the last step the
32. 9 Axle Sensor Setup Parameters Axle Sensor Spacings These parameters are normally set during the site installation and SHOULD NOT be casually changed Calibration factors can be entered using the Calibration Factors window These values are simply multiplied to the final weight computed from each sensor strip For example if it is set to 0 5 the weight computed would be halved a9 Calibration Factors O x Upstream Downstream Lane 1 W fos Save Lane 2 jos fos Exit The changes become effective only after the parameters are saved and then activated or when the Bulldog is restarted Calibration factors are applied as direct mutilplication to the original axle weights 4 Weights can be calibrated using speed ranges For example if the system tends to overestimate weights at a high speed it can be easily calibrated using the wizard shown below This window pops up when the menu item Speed Adjustment Factors is selected The entries represent the mid point of the speed range from which the rest of points are linearly interpolated B 1 a Weight Adjustment Factors per Speed El x Lane 1 0 gt 10 10 gt 20 20 gt 30 30 40 40 gt 50 50 gt 60 60 gt 70 70 gt 80 80 gt mph 1 3080 1 1366 1 0043 08396 0 8147 0 7444 0 6854 0 6349 0 5914 Lane 2 0 10 10 gt 20 20 gt 30 30 40 40 gt 50 50 gt 60 60 gt 70 70 80 80 gt mph 1 3080 1 1366 1 0043 0 8996 0 8147 0 7444 0 6854 0 6349
33. Figure 36 These wrinkles tend to flop when a wheel passes over causing generation of false axle signals It is important not to create the wrinkles during the installation but it is also recommended that an intelligent algorithm is developed to filter such false axle signals since careful installation is not always warranted This research used only flat bottom conveyer belts to construct weigh pads Some conveyer belts have horizontal or vertical grooves at the bottom These grooves are there to increase the friction against pulleys In the same way the grooved conveyer belts should increase the friction against the pavement surface and could provide a better fastening capability and stability However it is unclear whether it would help or harm the accuracy of the weight measurements An experimental study is recommended to test grooved weigh pads 60 References 1 S K Edward A M Clayton and R C Haas Evaluating pavement impacts of truck weight limits and enforcement levels Transportation Research Record No 1508 1995 2 AASHTO AASHTO Guide for Design of Pavement Structure American Association of State Highway and Transportation Officials Washington D C 1993 3 NCHRP 1 37A Using Mechanistic Principles to Implement Pavement Design National Cooperative Highway Research Program NCHRP Washington D C 2006 4 NCHRP 1 39 Traffic Data Collection Analysis and Forecasting for Mechanistic Pavement Design Nat
34. Proceedings of the Fifth IASTED International Conference on Circuits Signals and Systems CSS 2007 pp 233 238 Banff Canada July 2 4 2007 61 14 T Kwon and B Aryal Hardware in the loop simulator for weigh in motion system development environment Transportation Research Board 87 Annual Meeting Washington D C Jan 13 17 2008 15 B Aryal WIM development environment based on a hardware in loop simulator M S Thesis Department of Electrical Engineering University of Minnesota Duluth MN Aug 2007 16 A Safaai Jazi S A Ardekani and M Mehdikhani A Low Cost Fiber Optic Weigh in Motion Sensor SHRP ID UFR 90 002 National Research Council Washington D C 1990 17 M Bin and Z Xinguo Study of vehicle weigh in motion system based on fiber optic microbend sensor Proc of the International Conference on Intelligent Computation Technology and Automation ICICTA pp 458 461 May 2010 18 Measurement Specialties Inc Roadtrax BL piezoelectric axle sensor Product Description Measurement Specialties Inc Hampton VA Jan 2007 19 C Helg and L Pfohl Signal processing requirements for WIM LINEAS Type 9195 Kistler Instrumente AG Winterthur Switzerland 2000 20 Jay L Devore Probability and Statistics for Engineering and the Science 4 Ed Brooks Cole Publishing Company Pacific Grove CA 1995 21 T Kwon BullConverter User Manual Transportation Data
35. Research Laboratory University of Minnesota Duluth MN Aug 1 2012 62 Appendix A Weigh Pad Test Picture June 4 2010 Test at MnRoad The weigh pads are setup much like a pneumatic tube counter The sensors in the pictures are the first built weigh pads The left side two black strips are a pair of single lane single sensor weigh pads The right side wider strip is a single lane dual sensor weigh pad that contains two parallel BL sensor strips Aug 16 2011 MnRoad Demo Day About 25 people from MnDOT State Patrol Center for Transportation Studies and industry were invited for a weigh pad demonstration and presentations at the MnRoad facility The event started 10 00am and ended 2 30PM A 1 Nov 3 and 4 2011 Cotton TH 53 Test The top photograph shows an installation process of weigh pads on TH 53 A temporary traffic control truck was called in which can be seen in the back The bottom picture shows the removal process of weigh pads on the next day The weigh pad data was successfully collected for a side by side comparison with the IRD system in this s te A 2 A 3 Appendix B Weigh Pad System Setting Wizards To measure vehicle speeds spacing of axle sensors in each lane must be set As shown below the spacing can be set using both feet and or inches but the final value is always converted into feet and set Sensitivity for each sensor segment must be set which is supplied by the sensor manufacturer a
36. a The RMSE was computed for GVW speed and vehicle length or simply length defined by a summation of axle spacing i e the distance from the front axle to the last axle of a vehicle The results are e GVW RMSE 4 10696 Kips e Speed RMSE 1 28684 mph e Length RMSE 0 32029 feet According to this result GVW has the highest RMSE and length has the lowest RMSE This is expected due to its absolute range of values Since RMSE only represents differences of absolute values between two observations and cannot objectively compare different parameters a Normalized RMSE NRMSE is used in place of RMSE NRMSE essentially represents a percent error and is a better measure for comparison of different numerical range of parameters A commonly used NRMSE is obtained by dividing RMSE by the range of the observed values L e NRMSE __ AMSE Vimar Ymin 15 The NRMSEs computed according to Eq 15 are e GVW NRMSE 0 03880 3 88 e Speed NRMSE 0 02219 2 219 e Length NRMSE 0 00514 0 514 The percent differences between the IRD and weigh pad data are 4 for GVW 2 for speed 0 5 for length Based on this data it can be said that the two systems produced a very similar data GVW had the most difference and the vehicle lengths total axle spacing had the least difference 52 According to above data the weight data had most differences which is expected because weights are affected by several environmental factors independent
37. a heat gun and the voltage generated was measured using a voltmeter When a two feet segment of the BL sensor strip was heated to 200 F about 200 mV was produced even though no loads were applied When this heat generated signal was connected to a regular charge amp this signal was able to damage the field effect transistor FET of the input stage of a typical charge amp circuit Consequently the conditioning circuit of a charge amp must not only compensate for the heat effect but also should protect the input stage from a large flow of charges generated by heat The solution to the heat problem sought in this research was to design the circuit so that it quickly dissipates the heat generated charges before they damage the input stage op amps without affecting the charges generated by axle loads In order to design such a circuit temperature characteristics of pavement must be understood Pavement temperature is generally 15 affected by two factors the amount of sun radiation and air temperature One important property is that air temperature or the heat of pavement tends to change slowly in comparison to the load changes of moving vehicles More specifically axle loads of a moving vehicle on weigh pads change within tens of milliseconds while pavement temperature changes in a much slower rate such as tens of minutes The design of BL charge amp should utilize this discrepancy i e the heat effect is removed by adding a dissipation path Thi
38. a neta sika aa eee gs 29 32 2 OCHS Wizards Gt Seen ela OS Sic Ga les a Stes sea OEE al 33 3 2 3 Qutput Data Format na aee er e ia a E i AE E E t aa 34 Chapter 4 Weigh Pad Pavement Installation essssessecssocssooesoocesocessccssocesooesoocssseessosssoose 36 4 1 Installation on Highways DR Re R Ne e M 36 4 2 Air Cavity and Vibration Problems si man asa ami 39 Chapter 5 Experimental Results i 0scc cisccciceicsascicisosecescscsvsecsocesccseronsseedsesesoossepoonsavodecovacrseese 43 Sl I veh gy ce a 0 ge A i hoy ee e ERO Re ROS Pry er a Ge e ee 43 5 2 Experiments on Influence ofSpeedsonWeight 45 De TNC OLY oe ot oleh lr ln sal la wed na theca be ta aaa liei a aaa a aman ei 45 3 2 2 ALA Collection sesi istan akan a dayama omak sab odana lisans a n GR 46 kpek a Rv Mm YAZ EM o el TEN Ye ve ie Pel Sede YO 47 5 3 Side by Side Tests of Weigh PadVs IRDWIMSystems 49 5 3 1 Test Setup and Data Collection ismi tin ak l mlnami lella Beli eee 49 5 3 2 PEA Analys IS eke Se Revco hats ac TA Sah Sel Sl de aa CON Rae re eT hed YON 51 Chapter 6 Conclusions and Future Recommendations cssceecss0000s000000000000000000000000000000 59 tl 0 0 1 b 100 SM AZ EM e Valle ED eke Rn nt AM DE Ki RE ER e 59 6 2 Future Recommendations a i ie le im o la lala dalla elle 59 NM K Zr A SL KN rep 61 Appendix A Weigh Pad Test Pictures Appendix B Weigh Pad System Setting W
39. al ground since no loads are present and the axle signals should only appear on C2 and C3 However that was not the case Figure 39 shows a plot of the actual signals of all four channels Notice that CO has signals that appear for every axle signals of C2 and C3 even though no axle load was present To show that CO signals are false plot of CO line was disabled and the rest signals are shown in Figure 40 Notice that two axle signals from C2 and C3 are clearly visible as they are supposed to be and C1 signals remain close to ground This verifies that CO signals are the superfluous faulty signals that should not exist In order verify that if the air cavity indeed produced the unwanted signal one third of the air cavity was filled with glues and then tested again The magnitude of superfluous CO signals was significantly reduced This was encouraging so the air cavity was completely sealed and tested again The false signals did not completely disappear from CO but its magnitude was small enough to ignore i e it was less than the axle signal threshold This experimentally proves that air cavity in weigh pads can introduce faulty superfluous axle signals A question still remained is why CO has axle like spikes for every axle signals from other channels It is reasoned as follows It should be first noted that piezoelectric materials generate charge signals when loads acceleration are applied but also when vibration is present When a vehicle
40. ane AxleC Speed Axle Spacing feet Axle Weights Kips cl 2 5 77 Oo 11 41 31 11 41 31 11 41 27 11 41 27 11 41 23 11 41 23 11 41 19 11 41 19 11 41 15 11 41 15 10 3 6 2 16 4 5 1 10 05 9 28 8 50 7 73 10 04 64 20 5 6 01 10 01 7 10 3 6 2 16 4 5 1 10 04 9 28 8 50 7 73 10 05 64 20 6 6 01 10 03 7 10 3 6 2 16 4 5 1 10 05 9 28 8 50 7 73 10 05 64 20 5 6 01 10 00 56 8 2 1 76 1 76 7 10 3 6 2 16 4 5 1 10 05 9 27 8 51 7 73 10 05 64 20 5 6 03 10 01 56 8 2 1 76 1 76 aNEO ROD 0D a W AN AMAM MM AMM OM Oh N A O N aA ao Maoa lz Figure 29 GUI screen shot of the developed weigh pad console In the Figure 29 screen both lanes are set to a northbound along with the arrows indicating the traffic direction in reference to the console box When the arrow direction is changed drawing of the vehicle model automatically changes its heading to match up with the arrow direction There is also a Release button in the middle right This button toggles its states between Hold and Release If it is pressed when the button text is Hold the vehShow controls are immediately frozen holding the display of the last vehicle on that lane If the button is pressed when its text shows Relaese the vehShow control is released and turns back to a normal mode i e the display is continuously updated as a new vehicle arrives This Hold Release function is useful when there is a need to inspect detail
41. arison verified that the data quality difference between the portable on pavement and a permanent in pavement system is minute It should also be noted that the data downloaded from the weigh pad system is compatible with the data format required by the BullReport a standard WIM data tool used by MnDOT consequently the same data tool developed for in pavement systems can be reused for the portable weigh pad system With few improvements the researchers believe that the weigh pad system is a solution for bringing the WIM technology to local roads at a low cost 6 2 Future Recommendations The developed system is battery operated but it only lasts for about 25 hours In traffic data collection traffic engineers typically collect short duration counts for two to three days using portable traffic counters 9 The battery run time of a portable WIM system should at least support an equivalent duration to be acceptable as a practical tool Extending the battery run time can be accomplished in two ways 1 increase the battery capacity or 2 use a low energy circuit Application of both approaches along with cost optimization is recommended Currently the life of weigh pads is completely unknown Since the pad material is reinforced rubber it will wear out and will need to be replaced at some point in time The breaking point or useable life of the weigh pads may be expressed in terms of traffic volume or Equivalent Single Axle Load ESAL Whateve
42. ch a more realistic model was developed and used The closed form of the new axle signal model in a continuous form is shown next by two equations O ep 1 Ni f x zacix 2 In Eg 1 a is the peak of the Gaussian function b is the amount of shift in x axis and c controls the width of the signal This function is plotted in Figure 26 The function f x in Eq 1 has a close form solution for its integration and it is shown in Eq 2 Since the computational model 25 cannot use the signal support range from a negative to positive infinity some limit has to be applied The width of the signal support area is selected as 5c as shown in Figure 26 voltage LA gt x gt Time t Figure 26 Gaussian axle signal model 3 1 1 Signal Modeling and Digital Signal Generation This subsection shows an example of finding a and c of the model 1 given the complete list of parameters of an axle as shown below Axle Weight W Sampling Rate R Samples Sec Footprint Length feet Speed v foot sec Sensor width w feet Multiplication factor f ound ound Since the signal support area is the footprint of the tire and sensor width it can be expressed as Se LEW sec v Therefore c is computed as l w c 3 z 3 The time it takes to go through a single sensor width is t sec 4 v Let the area under the axle signal be A i e 26 Ay D f x dx acNn 5 Deriving from Eq 1
43. d a single lane right weigh pads 14 1 8 1 6 1 4 1 2 1 0 8 Volt 0 6 0 4 0 2 0 0 2 0 4 0 6 16 500 17 000 17 500 18 000 18 500 19 000 19 500 20 000 20 500 21 000 21 500 22 000 Sample Index Figure 12 WIM waveforms of a Toyota Sienna van captured from a pair of two lane single strip weigh pads 2 3 Charge Amplifier The RoadTrax BL sensor is a piezoelectric sensor that produces charge signals in response to acceleration or load The charge signals must be converted to voltage signals in order to be able to sample the values using an ADC The converter that converts from a charge signal to a voltage signal is called a charge amplifier or charge amp in short Presently charge amps for the BL sensor are not commercially available thus the research team had to design and build new charge amps for this project This section describes the design One of the challenges in developing a charge amp for BL sensors is that it generates charge signals in response to heat This heat sensitivity can be a serious problem in some regions For example in Arizona peak asphalt temperatures have been recorded up to 160 F 71 1 C in June and July Since weigh pads directly touch the pavement on installation the pavement heat is directly transferred to BL sensors which causes generation of a large amount of charge signals To find out the effect of heat on BL sensors in the lab the sensors were heated using
44. e WIM system over an in pavement permanent WIM station for certain cases First since portable WIM sensors are not installed by cutting pavements it does not weaken the pavement structure With this property a portable WIM system would be more favorable to be installed on structurally sensitive areas such as on bridge decks Second the measurement locations can be freely selected and moved This property could be used as a preliminary study to locate a permanent WIM site Third since a single system can be reused for many locations by moving around the deployment cost is very low Forth the maintenance cost is lower than that of permanent WIM stations because no electricity or communication link cost is required Consequently there are sufficient motivations as well as needs to develop a practical portable WIM system There are enormous challenges to develop a practical portable WIM system First sensors must be durable and have a strong enough grip on the pavement surface to hold against the traction forces of heavy trucks Second the sensors must be easy to install and remove from pavement These two factors are a kind of opposing conditions and difficult to be met at the same time More specifically if sensors are strongly fastened on the pavement they would produce more accurate readings because of less vibration but they would not be easily removable Regardless how good a portable installation would be it would not be as secure as th
45. e Digitized Signals Back to Weight for Verification From the discrete signals generated the signal area is computed by simply adding each sample point as shown in Eq 8 The signal area represents the weight and the actual weight can be computed using Eq 9 Computation of these two equations occurs in the weigh pad console 28 Me ER w v R Avg Topi Ay d A S pound v wR where v vehicle speed in foot sec w sensor width in feet R sampling rate in samples sec Example If the area of the signal generated for the given example is summed up Continuing from Example 1 the area under the axle waveform is now computed back to the corresponding weight It started with 20 000 pounds of axle load in Example 1 and should end up 20 000 pounds when it computes back Example 2 When the digital samples generated by Example 1 is added up the total is 96 74431 This sum represents the weight of the axle and should be converted to pounds using Eq 9 i e A 96 74431 0 164 t sec 88 _ 96 7443 1x 1541 4765 x 2 20 000 pounds pound t x 4000 p This example illustrates that the axle signal model defined in Eq 1 is correct In the WIM system programming Eqs 8 and 9 were used to compute the axle weights 3 2 System Software 3 2 1 Overall GUI and Tool Sets The console computer described in Section 2 5 runs on a Windows embedded XP OS and is equipped with an SVGA LCD monitor 800 x
46. e MnRoad facility Among them the forth one provided a relatively quick as well as secure installation thus it was chosen as the installation method for high speed highways in this project The road selected for testing highway traffic was the Minnesota Trunk Highway 53 TH 53 The posted speed limit of TH 53 at the test site is 65 mph 104 6 Km h but the majority 70 of vehicles drives between 70 75 mph 113 121 Km h The weigh pads were installed on the northbound two lanes of TH 53 at the Cotton WIM station A newly constructed weigh pad pair that does not have air cavity was installed MnDOT Office of Transportation Data amp Analysis MnDOT TDA ordered the lane closure and supplied the installation tools needed The three tools used are listed in Figure 33 Hammer drill with 1 4 diameter 6 long drill bits Strong bonding utility table black color black Gorilla tape was used Gm Sleeve Anchor diameter 2 1 4 length with washer produced by Red Head was used Figure 33 Tools needed for weigh pad installation 36 After traffic control has shut down the lane the weigh pad installation steps applied are as follows First a 1 4 diameter hole is drilled through the pad and pavement for a total depth of 2 4 using a hammer drill and then a sleeve anchor is inserted to the drilled hole after placing a 4 washer The screw on top of the sleeve anchor is turned clockwise t
47. e computational model and the overall software design at the system level 3 1 Axle Computational Model The hardware in loop HIL WIM signal simulator which was developed by the PI and his student in 2007 14 15 was extensively used in the software development phase of the weigh pad system The HIL simulator is a WIM hardware software hybrid simulator and can replace axle load and loop signals with software simulated electric voltage signals by passing the axle model generated values to a Digital to Analog Converter DAC It operates in real time and produces axle load waveforms based on a set of user defined inputs These include vehicles per minute mix of vehicle types speed range and definition of each vehicle axle load characteristics number of axles axle weights tire footprint lengths and axle spacing Axle signals of various traffic conditions and any mix of vehicle types can be generated in real time using the HIL simulator The HIL simulation allows for any real WIM system to be directly tested under various traffic conditions without installing sensors on the road In addition a number of faulty conditions of axle sensors can be simulated and tested using the HIL simulator 14 15 Figure 25 Hardware in Loop HIL WIM signal simulator The HIL simulator requires an axle load signal model to generate the axle load waveforms In the original HIL version a trapezoidal signal model was developed and used 14 In this resear
48. e final tests were conducted at an existing in pavement WIM site for a side by side comparison The chosen road was one of the Minnesota trunk highways and had an average traffic speed of about 67 mph The weigh pads were installed right next to a WIM site constructed using Kistler Lineas quartz sensors and an IRD iSync WIM system A total of 3 235 vehicle records were compared for three parameters GVW speed and axle spacing Normalized Root Mean Square Errors NRMSEs between two systems on GVW speed and axle spacing were 3 88 2 22 and 0 5 respectively Correlation coefficients between two systems for GVW speed and axle spacing were 0 97 0 97 and 0 99 respectively The coefficient of determinations denoted as R were 0 93 0 93 and 0 99 for GVW speed and axle spacing respectively Lastly the difference in vehicle classifications between the two systems was merely 1 5 All of the comparison measures indicate that the WIM data obtained by the weigh pad system is only a few percentage points different than the data of the same traffic obtained by a permanent in pavement WIM system This test result suggests that the weigh pad system developed in this project provides WIM data with a quality similar to that of a permanent in pavement WIM system In conclusion this project successfully demonstrated that a reusable portable WIM system that would work much like a pneumatic tube counter can be built and deployed A side by side compariso
49. e sensors of an equivalent permanent WIM station that are installed inside the pavement Consequently accuracy of a portable WIM system must be achieved from signals generated by obscure sensor installations Each passage of a heavy truck may slightly move or vibrate the sensors which would create superfluous signals In order to filter out these unwanted signals an intelligent algorithm that can isolate faulty forces from the main load force must be developed which would not be simple In order to successfully meet the mentioned challenges it is important to have working knowledge on WIM sensors and axle forces understand materials and pavement properties and have experienced in processing of real time WIM signals Much of such information is not available in the literature because WIM systems are mostly developed from a proprietary environment The PI Principal Investigator and MnDOT have been involved in several WIM system development projects for a number of years The following paragraph describes the past related research and development R amp D efforts In 2003 2004 the PI developed a signal probe for Kistler Lineas sensors as one of the sponsored projects The result was presented at the NATMEC 2004 conference 10 and in the project report 11 This system is equipped with charge amps and provides real time plots of axle load waveforms from the charge signals of Lineas Quartz WIM sensors At the same time it can record the raw axl
50. e the charge signals generated by the heat Since the charge amp was designed to remove any slowly changing signals a static weight test was conducted to verify the signal response to an unchanging weight About 60 pounds of weight was placed on top of the test BL sensor and then the charge amp output was observed The voltage level rose when the weight was initially placed on the BL sensor but it was gradually decreased to the signal ground as the weight becomes static and initially generated charges are depleted The charge amp responded as expected to the heat and static weights After confirming the operation of charge amps in the lab the next step was to test its waveforms on moving vehicles Low speed tests were done at one of the UMD parking lots while high speed tests were conducted at the MnRoad facility The charge amps responded well for the slow speed tests in parking lots producing reasonable axle load waveforms Many driving tests were conducted at MnRoad and the two test results shown in Figures 41 and 42 are next described For sampling of the charge amp output signals a PCI based ADC MCC PCI DAS6013 was used in both cases The voltage signals were sampled at 4 096 samples per second Figure 41 shows a waveform of a two axle vehicle produced by a two channel charge amp The test vehicle in this case was a Toyota Siena van with the GVW 4 600 pounds The waveform shape and size was within the expected range Next a 5 axle semi traile
51. e waveforms in a binary form for future reviews This system was designed as a portable diagnostic tool to be used in the field for testing Kistler Lineas sensors Field tests revealed several abnormal conditions of charge signals 13 In 2006 the PI and his graduate students successfully completed development of an eight channel real time WIM system based on a PC and off the shelf components and installed as a working WIM station 12 This real time WIM system allows the users to examine the raw axle signals without removing the sensor connections from the WIM system i e it combines the WIM probe ideas with a regular WIM system This project demonstrated that a WIM system can be easily built using off the shelf components During this project new signal processing techniques and signal modeling for WIM systems were developed and published in 13 Another innovation created during this project was development of a hardware in the loop HIL simulator for generating real time axle load waveforms in voltages using mathematical axle models 14 15 The HIL simulator can generate ideal as well as faulty axle signals consequently the WIM systems can be tested under simulated traffic conditions The HIL simulator was a critical tool for developing the PC based WIM system The HIL simulator was also extensively used during the software development phase of the new portable WIM system saving a huge amount of development time In the past most R amp D e
52. effect would not diminish Collection of more data is recommended to finalize the speed effects Table 8 Log Calibration Factors speed 5 mph_ 15 mph 25 mph 35 mph 45 mph 55 mph 65 mph 75 mph 85 mph calfac 1 7692 1 2959 0 9925 0 8598 0 7818 0 7290 0 6902 0 6600 0 6358 y e 2 2045 12481 2 2065 weight Kips 0 10 90 speed mph Figure 46 Log regression of weight data by different speeds 5 3 Side by Side Tests of Weigh Pad Vs IRD WIM Systems 5 3 1 Test Setup and Data Collection It is interesting to compare how the weigh pad based portable WIM system performs against an in pavement installation such as the Kistler Lineas Quartz sensors with an IRD iSync WIM 49 system most of MnDOT WIM systems One of the simplest ways to compare two systems would be a side by side comparison at the same location for the same traffic by installing the weigh pads side by side along with the Kistler quartz sensors The location selected for the side by side comparison was the Cotton WIM site on the Minnesota Trunk Highway 53 TH 53 at the mile point 42 The posted speed limit of TH 53 at the test site is 65 mph but the majority 70 of vehicles drives between 70 75 mph This WIM site has four lanes in which Kistler Lineas quartz sensors are installed and operational An IRD iSync WIM system is available in the roadside cabinet which collects the WIM data of this site
53. en though they are thin less than 1 3 inches in the center and the edges are tapered to a gradual slope When a vehicle drives over the weigh pads on a high speed a sound of hitting a small bump can be clearly heard This bumping sound becomes louder as the vehicle speed increases This begs the question Does the vehicle speed affect the vehicle weight measured by the weigh pad This section describes a brief theoretical analysis and the experimental results To analyze the loading of an extruded weigh pad sensor we consider a load axle sensor installed on a slope of 4 as opposed to installed on a flat surface A slope installation and the related forces are illustrated in Figure 43 In the diagram the little rectangle on the slope is the axle load sensor If the slope is zero right side diagram the sensor load due to the vehicle s horizontal force becomes zero because F sin 0 0 thus the load received by the sensor Fs is equal to the gravitational force only i e the weight F On the other hand if the senor is installed on a slope as shown in the left side of the diagram in Figure 43 the total force received by the sensor is the summation of the horizontal force generated by the acceleration and the weight of the vehicle by gravity i e F F F 10 45 In theory as the vehicle s speed increases the horizontal force Fn applied to the load sensor should increase while the gravitational weight of the vehicle remai
54. ent To examine the pavement heat effect the weigh pads and charge amps were tested under the pavement temperature range of 85 134 F This pavement temperature range was available at MnRoad in the summer of 2011 in Minnesota The idle levels of the signal observed for the entire temperature range remained close to the signal ground The voltage level movement was almost none existent in comparison to the lab tests which applied the temperatures up to 350 F using a heat gun The effects to the axle waveforms by the tested temperature range 85 134 F were insignificant or negligible In summary the charge amp circuit was designed to remove any signals that change slower than the period defined by the designed time constant of 20 seconds The next question is then Does this slow signal removal adversely affect the fast changing axle load signals According to many observations and example waveforms like the ones shown in Figures 41 and 42 it did not affect the actual axle signals because most axles move much faster than the 20 second time constant for traveling the sensor width of 0 7 mm this translates to an equivalent speed of 0 000747 mph Pavement temperatures on the other hand slowly change such as the rate in the order of 10s of minutes which triggers removal of the signals 5 2 Experiments on Influence of Speeds on Weight 5 2 1 Theory Because weigh pads are fastened on the surface of the pavement they are extruded ev
55. erage GVW Ratio over GVW Ranges in Kips and Speed Rangesinmph 56 Table 13 NumberofVehicleRecordsintheDefinedRange 57 Executive Summary Weigh in Motion WIM systems produce individual vehicle records of traffic information that includes lane number time stamp speed axle loads axle spacing and classification of the vehicle type This detailed traffic information has been used in a wide range of applications i e pavement analysis and design overweight enforcements traffic data analysis and reporting freight estimation traffic monitoring etc Although benefits of WIM data are evident initial construction and the subsequent maintenance of permanent roadside WIM stations are expensive WIM stations therefore have mainly been installed on roadways with heavy traffic such as interstate and trunk highways They are almost nonexistent on rural local roads because of low Average Daily Traffic ADT and difficulty of cost justification However low ADT on rural roads does not mean fewer overweight violations or diminish needs for protecting the roads from overweight vehicles Heavy truck volumes on local roads indeed have been increasing caused by higher demands on agricultural commodities This raises a grave concern for many local transportation engineers because it could significantly shorten the life of local roadways To monitor road wear or to protect from overweight vehicles
56. esults and analysis Chapter 6 concludes the report with final remarks and future recommendations Chapter 2 Hardware Design 2 1 Overall System The hardware portion of the weigh pad system consists of four modules axle weight sensors charge amplifiers amps Analog to Digital Converter ADC and a computing unit A block diagram of these modules along with their signal flow is illustrated in Figure 1 This chapter describes details of the design and implementation of each module Axle i N Charge A D N Computing ae LS Amplifier Converter Unit Figure 1 Hardware block diagram of the weigh pad WIM system 2 2 Axle Weight Sensors and Weigh Pad Design 2 2 1 Axle Sensor and the Pad Material Axle load sensors must produce measurable electrical signals in response to axle loads Since the objective of this project was to develop a portable WIM system that can be installed much like a pneumatic tube counter thin and flat sensor strips are most desirable The sensor that satisfies these two conditions was the Roadtrax BL Brass Linguini Piezoelectric Axle Sensor or simply BL sensor made by Measurement Specialties Inc 18 Specifications of the BL sensor supplied by the manufacturer are summarized in Table 1 source from 18 The sensor length chosen was 12 feet to match with the standard U S traffic lane width The length of the coaxial cable attached to the sensor can be
57. f a pair of weigh pads and a console computer For installation two weigh pads are laid across the traffic lane separated by a known distance typically 12 to 16 ft and fastened on the pavement surface using sleeve anchor screws The edges are then taped using strong bonding utility tapes Since on pavement installations produce much stronger charge signals than in pavement installations a customized charge amp was developed to handle the different charge responses In addition a durable field ready enclosure that houses all necessary electronic components and a computing unit was designed and fabricated The final system consists of two parts a console box and a pair of weigh pads and is truly portable One of the advantages of the weigh pad system is that sensor installation does not cut into the pavement Since the installation does not weaken the pavement structure it would be safe to use on structurally sensitive areas such as on bridge decks To verify WIM capabilities of the weigh pad system driving tests were conducted at the MnRoad facility and also on Minnesota Trunk Highway 53 TH 53 Two types of effects were tested at MnRoad which are the effects of temperature and speed to the weight measurements For temperature tests a single test vehicle with a known weight was driven over the weigh pads repeatedly in the pavement temperature range 85 135 F and the corresponding gross vehicle weights GVWs translated from the axle wavef
58. fforts on WIM systems have been given to in pavement permanent WIM stations Little efforts have been given to development of portable WIM systems One early research that was showing a promise as a portable WIM system was fiber optic sensors 16 17 The pressure on the sensor causes optical fiber deformed which leads to the loss of output light The vehicle weight is obtained through measuring the variation of light intensity in optical fiber 2 Since optical fibers are very thin they have an attractive physical form factor for developing a portable WIM sensor Although this line of research has been commercialized for in pavement installations it has yet to evolve to a portable WIM system This research utilizes a piezoelectric load sensing technology because it is still the most widely used load sensing technology Some of piezoelectric strips are now specifically designed for WIM applications and readily available 18 19 For example the quartz piezoelectric sensors developed by Kistler Inc 19 have been widely accepted for in pavement implementation by many states and successfully used for collecting WIM data for many years For this project the research team decided to use the RoadTrax BL Brass Linguini sensors referred to as BL sensors 18 which are piezoelectric strips designed for WIM applications classified as Class 1 or WIM applications according to the manufacturer s classification The BL sensor strips are thin l
59. inition file It accepts either a metric tym or English unit definition file 33 tye in the same way the BullConverter reads in The last important entry is setting up where to store the data in the file system The user must set the root directory of the WIM data to be stored using the Browse button in the GroupBox named WIM Data Root Folder The data is then stored in a subdirectory named with a yyyymmdd format If the subdirectory does not exist the software will automatically create the directory The user is only responsible to set the root folder for the WIM data As the last step the user must press the Save button to save all entries If the Exit button is pressed without saving all of the new entries are not updated and the old entries will be remained It should be noted that the setup window only serves as a GUI for users and the values are not actually stored in the window but in a Settings ini file al Site Setup a x r Site ID 3digit num 031 Location Two Harbors HW61 mp 16 4 Direction Arrow Display Lane 1 N v lt X Lane 2 N X eas r Veh Class Definition File tye or tym File Path C WIM_development WeighPad 1 2 bin min5 tye WIM Data Root Fol DataPath C wWIM_development WeighPad 1 2 WIMdata Browse All information in this setup is activated when Bulldog is restarted Save Exit Figure 32 Site setup window 3 2 3 Output Data Format
60. ional Cooperative Highway Research Program NCHRP Washington D C 2003 5 ASTM 1318 02 Standard Specification for Highway Weigh In Motion WIM Systems and User Requirements and Test Methods American Society for Testing and Materials ASTM West Conshohoken PA 2002 6 Steve Jessberger Understanding traffic inputs for the pavement design guide North American Travel Monitoring Exposition and Conference NATMEC Loews Coronado Bay San Diego CA 2004 7 A Papagiannakis M Bracher J Li and N Jackson Traffic load data requirements for pavement design 6 International Conference on Managing Pavements Brisbane Australia 2004 8 Y H Huang Pavement Analysis and Design 2 Ed Pearson Prentice Hall Upper Saddle River NJ 2004 9 FHWA Traffic Monitoring Guide U S Department of Transportation Office of Highway Policy Information Washington D C May 2001 10 T Kwon Signal probe and processing methods for improving WIM data North American Travel Monitoring Exposition and Conference NATMEC June 27 30 2004 Loews Coronado Bay San Diego CA 11 T Kwon Annual Report Transportation Data Research Laboratory 2004 CTS 06 03 80 pages Minneapolis MN Apr 2006 12 T Kwon and B Aryal Development of a PC Based Eight Channel WIM System Minnesota Department of Transportation St Paul MN Oct 2007 13 T Kwon Signal processing of piezoelectric weigh in motion systems
61. it is necessary to adjust back to a slower rate than the log function for below 10 mph A piecewise linear function for a practical implementation of speed calibration could be used to allow implementation of any non linear mapping relationships This approach was used in the weigh pad system implementation and the user can enter calfac in 5 mph 8 Km h spacing up to 90 mph 145 Km h In summary the theory presented in the Sub Section 5 2 1 and the test results from MnRoad show that speed of a vehicle influences the total force applied to the weigh pad axle sensor However there must be a caution in interpreting this data The obtained data may be influenced 48 also by installation variances and the associated vibration If weigh pads are not securely fastened to the pavement the pads vibrate when a tire hits the sensor As the vehicle increases the speed the amount of this vibration also increases Since piezoelectric sensors produce higher charge signals in response to vibrations another factor influencing on top of the speed might be vibration More specifically vibration effect is amplified by higher speed resulting stronger charge signals What this suggests is that the weight vs speed relationship derived through the speed data may be an overestimate and the weight values would be reduced if the sensor pads are more securely fastened to the pavement The vibration effect could be diminished if installation is ideal but the slope
62. ity generated noise on Channel 0 C0 Volt 14 000 14 500 15 000 15 500 16 000 16 500 17 000 17 500 18 000 Sample Index Figure 40 Removing Channel 0 signal clears the superfluous signals 42 Chapter 5 Experimental Results 5 1 Charge Amp Tests Charge amplifiers amps are a crucial part of any WIM system and must be carefully designed and tested The charge amps for this project were custom designed and fully described in Section 2 3 The initial Printed Circuit Board PCB prototypes for building charge amps were produced using an LPKF machine PCB milling machine available in one of the research labs at UMD The charge amp built using the LPKF machine can be seen in Figure 15 The final PCBs were commercial grade boards produced in a factory through one of the PCB prototyping services and they can be seen from Figures 14 and 17 One of the early tests of the charge amps was a heat test The input of the prototype charge amp was connected to a BL sensor while the output was connected to an oscilloscope The BL sensor was heated gradually up to 350 F using a hot air heat gun and the output of the charge amp was monitored The voltage level was initially rising but constantly pulled down to the signal ground level as soon as the heat gun was removed In order to hold the charge amp output at a certain voltage level the heat had to be constantly applied at an increasing rate This indicates that the charge amp was able to remov
63. izards Appendix C Sample Weigh Pad WIM Data List of Figures Figure 1 Hardware block diagram of the weigh pad WIM system 5 Figure 2 Dimensions of single lane sing le stripweigh pad 8 Figure 3 Asingle lane single stripweigh padprototype 8 Figure 4 Dimensions of single lane dual strip weigh pad 9 Figure5 Asingle lane dual stripweigh padprototype 10 Figure 6 A waveform generated by a single lane single strip weigh pad for a Toyota van 11 Figure 7 Five axle semi trailer truck WIM signal recorded from a single lane dual strip weigh DAG ZE EM e Ee ath e cet AR N PYAR e apne PAY an tm Ne 12 Figure 8 Dimensions of the Two lane single stripweigh pad topview 13 Figure 9 Prototypes ofapairoftwo lane single stripweigh pads 13 Figure 10 Two lane single strip weigh pads installed ontheMnRoadtestsite 14 Figure 11 Comparison of a two lane left and a single lane right weigh padS 14 Figure 12 WIM waveforms of a Toyota Sienna van captured from a pair of two lane single strip Bd 18 e e NR ZN e e A DM 15 Fig r 13 Basic charge anip cir Curt a al ala alakaya aa ak asamamada laa yalak aa 17 Figure 14 A prototype two channelchargeampbuil
64. levard MS 330 St Paul MN 55155 15 Supplementary Notes http www lrrb org pdf 201238 pdf 16 Abstract Limit 250 words Installing permanent in pavement weigh in motion WIM stations on local roads is very expensive and requires recurring costs of maintenance trips electricity and communication For county roads with limited average daily traffic ADT volume such a high cost of installation and maintenance is rarely justifiable One solution to bring WIM technologies to local roads is to utilize a portable WIM system much like pneumatic tube counters used in short duration traffic counts That is a single unit is reused in multiple locations for few days at a time This way WIM data is obtained without the cost of permanent in pavement WIM stations This report describes the results of a two year research project sponsored by the Minnesota Department of Transportation MnDOT to develop a portable WIM system that can be readily deployed on local roads The objective of this project was to develop a portable WIM system that would be used much like a pneumatic tube counter The developed system is battery operated low cost portable and easily installable on both rigid and flexible pavements The report includes a side by side comparison of data between the developed on pavement portable WIM system and an in pavement permanent WIM system 17 Document Analysis Descriptors 18 Availability Statement Weigh in motion Portable WIM system
65. messages it was found that the erroneous records were mainly due to vibrations generated by the air gap of the wrinkles in the sensor pad Figure 36 Flapping of the pad edges created false axle signals which led to an error message that indicates mismatch of left and right axle counts The weigh pad console presently does not filter false axle signals and it simply gives an error message without computing the axle weights This experience clearly taught us that it is extremely important to minimize or eliminate wrinkles during the weigh pad installation Moreover a 51 filtering algorithm that could identify and remove false axle signals is needed asa part of the weigh pad signal processing This would be one of the future recommended works Since the bad records cannot be used for a meaningful comparison against the IRD vehicle records all of the 587 bad vehicle records were removed from the weigh pad data The remaining 3 235 vehicle records were used for analysis In the subseguent analysis the weigh pad vehicle records referred would be the error free records with the raw data recorded using the default setup shown in Tables 9 and 10 If the data was calibrated afterward it will be stated In order to measure the difference between the IRD and Weigh Pad vehicle records a Root Mean Sguare Error RMSE is used as the first similarity test The RMSE is defined by RMSE 14 where x is the IRD data and y is the weigh pad dat
66. n mining are flexible durable and thin all of which are good properties for embedding the BL sensor strips into the material The flat and thin shape of conveyer belts allows easy assemble of sensor pads because weigh pads can be simply built by placing and gluing a RoadTrax BL sensor strip between two conveyer belts There are hundreds of different types of conveyer belts available for many different applications Conveyer belts with the product code 908860 made by Forbo Movement Systems were selected as the final pad material after consulting with the conveyer belt experts The specifications of the conveyer belt material selected are summarized in Table 2 Notice from the table that this material can handle heat up to 225 F 100 C This property is important because pavement temperatures in summer months can reach as high as 180 F 42 C in Arizona The color of the material is black and would blend well with pavement The material is durable but can be sanded to produce smooth edges The measured thickness of a single 908860 conveyer belt was about 0 1476 in 3 75 mm When two 908860 belts are sandwiched together the thickness was 0 295 in 7 5 mm This would be the center and the thickest part of the weigh pad The leading and trailing edges of the weigh pads are sanded down to about mm to reduce the bumpiness Table 2 Weigh Pad Material Specifications Product Model UTILITY 2 160 GRADE II 1 32X1 32 NA Transtex Product Code 908
67. n the previous WIM system development projects This board ran multiple years without an error Figure 18 PCI DAS6013 ADC board 19 2 5 Console Box and Enclosure A console box or computer in this report refers to a single enclosure that houses all necessary components of the stand alone weigh pad WIM system excluding the sensor pads With this console definition the weigh pad system simply consists of weigh pad sensors and a console box The console box includes a mother board a keyboard with a mouse pad charge amps A D converters an LCD monitor batteries and a charger At the beginning of this project a laptop computer was going to be used as the computing unit of the console since it is already eguipped with a display a large amount of data storage a keyboard with a mouse and I O interface ports However the laptop idea was quickly abandoned mainly due to the battery capacity that must be able to continuously support a minimum of 24 hours without charging The problem with laptop computers is that they are not energy efficient mostly lasting only up to six to eight hours In addition it is not simple to modify the battery management module embedded in a laptop computer to accept a large capacity external battery bank Another problem experienced is that there is no hard switch that can completely shut off the LCD monitor which would have saved a significant amount of energy since the LCD monitor is no longer needed after
68. n verified that the data quality difference between the portable on pavement and a permanent in pavement system is minute With few improvements the researchers believe that the weigh pad system is a solution for bringing the WIM technology to local roads at a low cost Chapter 1 Introduction A linear increase in load is known to have a forth power exponential increase in the acceleration of road wear which has been the basis for pavement design and maintenance for many years 1 Weigh in Motion WIM systems provide this vital traffic load data as the inputs for pavement design and management 2 7 In the NCHRP 2002 Mechanistic Empirical Design Guide 3 which is simply referred to as the 2002 Guide traffic is handled in terms of annual load distribution spectra by axle configuration The full spectra for single tandem tridem and quadrem axles are directly used as the design inputs On the other hand the traditional Equivalent Single Axle Load ESAL 8 which represents damage to pavement is still popularly used for pavement designs by many transportation departments Regardless which method is used WIM systems provide essential traffic load information for multiple applications WIM data also meets standard traffic monitoring needs 9 Although there are many benefits infrastructure cost for building permanent roadside WIM stations is expensive For example installing a WIM station for a four lane highway typically costs over 220
69. nd guidance throughout the project period The author also would like to thank Tim Clyne at the MnRoad facility for allowing the research team to use the low volume roads for controlled driving tests Table of Contents Ch pter 1 Ttroducllom dek ces eesn wc iscsi Seal aladne aba a AAA da be o aa dala Ke elm edrikana pia 1 Chapter 2 Hardware Design i e pasadese edindik s ne e dek din cat capscavescivcasavens cavvecseuevuscpngvescnsnave 5 Zelve li SVS CCIM a akape ala akim ale asal iile pe mai 5 2 2 Axle WeightSensorsandWeigh PadDesign 5 2 2 1 Axle Sensor and the Pad Material e ki e Rl aa al ee een lak 5 2 2 2 Single Lane Weigh Pads Design eid a tan kask da kill a ikti arak 7 2 2 3 Tw L ne Weieh Pad DE K ADR a aA 12 2 3 Charge A tp teh morse oeron nan ie dll date Gass wets bu be eae vale 15 2 4 Analog to Digital Converter CADC asa od Sous in mani Say nul ds oss 17 2 5 Console Box and Enclosure al elen ak m i gala e sl 20 Chapter 3 Weigh Pad System Software Design ccsccsssccsssssccssssscssssccsssccssssssssssseseees 25 3 1 Axie Computational Medela a a Aa A Aiko e ASR NA Ma ese 23 3 1 1 Signal Modeling and Digital Signal Generation 26 3 1 2 Computing the Digitized Signals Back to Weight for Verification 28 Deo SY SUSI SOUT Wye ce KR tas gee arcsec ENEA geo ues ne Pag ee e e AY 29 32 1 Overall GUL Gnd Tool Sets samimi ayla a
70. no damage was found from the sensor pads or the controller The console battery was able to support more than continuous 24 hours of operation Table 9 Setup Parameters Setup Parameter Lane Value Sensor Strip Spacing Lane 1 14 ft Lane 2 14 ft Sensitivity Lane 1 Upstream 6 25 Downstream 6 25 Lane 2 Upstream 6 25 Downstream 6 25 Calibration Factor Lane 1 Upstream 1 Downstream 1 Lane 2 Upstream 1 Downstream 1 Table 10 Limit Parameters Parameter Value Maximum Vehicle Length Possible 75 ft Maximum Axle Spacing Possible 4S ft Maximum Tire Footprint Length Possible 2 ft Minimum Possible Vehicle Speed 15 mph Maximum Possible Vehicle Speed 110 mph Axle Signal Threshold 0 12V for All 5 3 2 Data Analysis The data was collected after manually adjusting the weigh pad console clock to closely match with the IRD s system clock at the Cotton station The time synchronization made easy to locate the matching vehicle records between the two system outputs In the recorded data exactly 17 seconds were different between the two system s vehicle records for the entire data collection period This time difference was caused by manual clock synchronization of the two systems and not trying hard enough to match up to the second In the weigh pad system data 587 out of 3 822 vehicle records had errors 15 After looking through the data and error
71. ns the same This means that the force of the same vehicle on a higher speed would result in applying a higher load to the sensor when the sensor is on a slope This relationship was tested in this experiment Fg Fs Figure 43 Force and slope relationship 5 2 2 Data Collection To verify the theory shown in Section 5 2 1 weigh pads were set up at the low volume road of the MnRoad facility The Toyota Siena van shown in Figure 41 was again used as the test vehicle The static weight of the test vehicle was approximately 4 6 Kips 2 087 Kg This test road is flat and can be driven above 80 mph 129 Km h without interference or dangers associated with other traffic The setup is shown in Figure 44 The system includes a pavement temperature sensor in order to ensure that the data is collected within a small range of temperature variance The weigh pads were fastened on the pavement surface using a carpet tape at the bottom and Gorilla tapes high strength duct tape on both leading and trailing edges of the sensor pads The weigh pads used were newly constructed two lane weigh pads that had no air cavity The test date was Oct 16 2011 and the weather was clear The pavement temperature fluctuated between 66 75 F 19 24 C during the measurement period which was 2 25pm 4 41pm The wind was blowing at 20 30 mph 32 48 Km h east The test vehicle was driven multiple times at speeds close to 10 20 30 40 50 60 70 and 80 mph
72. o expand the sleeve which fastens the washer and thus the pad to pavement Sleeve anchors are installed by spacing approximately 2 feet apart reguiring 24 sleeve anchors for installing a pair of weigh pads After fastening the pads by sleeve anchors the leading and trailing edges of the pads are taped using a strong bonding utility tape such as a Gorilla tape shown in Figure 33 Taping prevents the sensor edges lifted by air drag or rolling resistance generated by wheels traveling through the weigh pad The taping also reduces trapping of air underneath the pads Although it was not used in this installation application of a carpet tape or any double sided tape underneath the sensor pads would help to further fasten the pads to the pavement For the Cotton TH 53 tests double sided carpet tapes were not used Two lane installation can be done by closing one lane at a time The sensor pads are installed by first unrolling the initial half of the roll and then fastening the pad to the pavement as shown in Figure 34 While the crew was installing the weigh pads the install time was measured The total time that took to install two 24 ft 7 32 m weigh pads for the two lane highway took about 30 minutes or 15 minutes per lane The crew mentioned that 20 minutes would be sufficient for two lane installation if it was not the first time installation To be conservative the PI believes that allocating 15 minutes per lane would be a good estimate for ins
73. oice Three USB based data acquisition products were purchased and tested Two basic ADC requirements were set 1 all boards must have 16 bit resolution in the analog to digital conversion and 2 at least 100K samples per second S s sampling rate must be supported The three boards purchased and tested include USB 7202 8 channels with 16 bit resolution at max 200K S s made by Measurement Computing Inc USB AI16 16A 16 channels with 16 bit resolution at 500K S s made by Access I O Products Inc and USB 6210 16 channels with 16 bit resolution at 200K S s by the National Instruments Inc 17 AlI three boards were programmed and tested for the performance and software development efficiency using a laptop computer Figures 15 16 and 17 show the test setup of the three different boards mentioned In the case of the USB 7202 board data over run errors were frequently observed resulting in much less than the claimed 100K S s sampling rate The cause of less than specified sampling rate was found to be caused by the inefficient software driver The next board tested was the USB AI16 16A This board had a much better real time data rate than the USB 7202 board but the software development DLL tool was hard to use In addition the connectors occupied too much space for a portable enclosure Lastly the USB 6210 was tested The software development environment called the NI Measurement Studio included many examples and made the software development
74. ong and strong which are the desirable properties for developing portable WIM sensors One of the objectives of this project was developing a portable WIM system that would be used much like a pneumatic tube counter In order to accomplish that goal the BL sensor strips were sandwiched and glued between two thin conveyer belts Conveyer belts provide flexibility and durability needed for on pavement installations This new sensors were called weigh pads A standard weigh pad has a length of 24 ft covering two lanes and a width of 1 ft The thickest part of the pad is in the middle and only 0 3 in The leading and trailing edges of weigh pads are sanded off to create a smooth rising and falling slopes respectively It can be easily wrapped around as a roll about 19 in diameter i e packed like a roll of pneumatic tubes Since the thickest part of the pad is only 0 3 in and the edges are smooth motorists of the traffic in general do not feel any bump from the pads installed on the roads For installation sensors are laid across the traffic lane and fastened on the pavement using sleeve anchor screws In addition the edges are taped using strong bonding utility tapes Piezoelectric sensors produce charges in response to loads and a circuit must convert the charge into a voltage A circuit that can convert from charges to voltage is called a charge amp One of the challenges of designing a charge amp for BL sensors is that they are sensitive to hea
75. ordered between 100 300 feet A 100 feet coaxial cable was used for the prototype The product data did not specify the thickness neither the width of the BL sensor so both values were manually measured in the lab The measured thickness of the sensor strip was 0 07874 in 2 00 mm while the width was 0 263 in 6 68 mm Table 1 RoadTrax BL Sensor Specifications Sensor Model RoadTrax BL Traffic Sensor Sensor Length 12 ft Sensor Thickness 0 07874 in Sensor Width 0 263 in Capacitance 10 67 pF Cable Length 100 300 ft Cable Type RG 58 with burial rated Capacitance of Cable 8 05 nF lt C lt 14 50 nF Dissipation 0 0294 Average Sensitivity 49 pC N Material Uniformity 7 Weight 3 pounds Part Number 6 1005438 1 The next consideration is finding a strong material that can effectively embed the sensor strips and coaxial cables to protect from the abusive loads of highway traffic Initially tested heavy duty materials include 1050 Ballistic nylon Toughtek Neoprene fabric Toughtek non slip rubberized mesh and textured Neoprene rubber sheets However none of these materials met the three requirements the research team was looking for which are durability flexibility and manufacturability from the University lab After searching through a number of different materials the research team eventually decided to use industrial conveyer belts It was learned that the type of conveyer belts used i
76. orms were analyzed The expected trend was for the GVW to increase as the pavement temperature rose since heat increases charge production of the piezoelectric sensors but the data did not show any trend This outcome is mainly attributed to the charge amp design in which it filters out any signal components that have a period longer than 20 sec Pavement temperatures in general change over a longer time period such as in the order 10s of minutes which are clearly outside the 20 sec time constant Consequently most charge signals generated by the pavement heat must have been drained from the charge amp The next test was speed effect on vehicle weight Since weigh pads are installed on the surface of the pavement by fastening the pads they are slightly extruded When a vehicle drives over the installed weigh pads a sound of hitting a small bump can be clearly heard This bumping sound becomes louder as the vehicle speed increases This begs the question Does the vehicle speed affect the vehicle weight measurements To answer this question the same test vehicle was driven multiple times at speeds close to 10 20 30 40 50 60 70 and 80 mph and the corresponding weights were analyzed The data showed an increasing trend of weights as the speed increased This result explains the bigger bumping sound as the vehicle speed increases and suggests that there is a need for a calibration of the measured weight with respect to the vehicle speed Th
77. ors The first two BNC connectors from the right edge are for the first lane the first BNC for upstream and the second for downstream The last two BNC connectors are for the second lane In the low right side corner of the back side an input port for thermocouple type k probe can be seen Figure 24 This is to measure the pavement temperature which might be needed in the future to compensate the weight values with respect to the pavement temperature The enclosure is built using 25 thick aluminum sheets thus it is exceptionally strong The user may actually sit on it while he or she is working in the field The lid can be locked using a padlock A drawback of the tough enclosure was its weight When all of the components are enclosed the console box weighed about 35 5 pounds 16 1 Kg which was slightly heavier than originally expected However this weight should still be acceptable for most people to carry around in the field Figure 20 Console computer enclosure specification 22 Figure 21 Custom enclosure built using aluminum sheets Figure 22 Middle layer of the console computer 23 Figure 23 Top layer of the console computer Figure 24 Back side of the console computer enclosure 24 Chapter 3 o Weigh Pad System Software Design The weigh pad WIM system may be divided into two parts hardware and software This chapter describes the software part of the system and it includes description of th
78. pad width to 12 in 30 5 cm from 6 in 15 2 cm The final weigh pad length was 25 5 ft 7 7 m which is 24 ft 7 3 m sensor length plus 1 5 ft 0 46 m extra length for protecting the connector of lead wires and a flap for screw installation The final design is shown in Figure 8 and the prototype weigh pads constructed according to the specification are shown in Figure 9 Figure 10 shows a pair of two lane single strip weigh pads installed for a test road Figure 11 shows a visual comparison of a two lane weigh pad roll left against a single lane weigh pad roll right In a ballpark figure a two lane weigh pad is 24 x 1 ft 7 3 x 0 3 m and a single lane weigh pad is 12 x 0 5 ft 3 7 x 0 15 m Figure 12 shows axle load waveforms of a Toyota Siena van generated by a pair of two lane weigh pads In the graph channel 0 CO is the signals from the leading upstream sensor strip and channel 1 C1 is the signals from the trailing downstream sensor strip Channels C2 and C3 are connected to the far lane sensor strips where no axle loads are present in this example 12 SS Coax Cables Sensor Strips Br ee ne LL Figure 8 Dimensions of the Two lane single strip weigh pad top view 13 Figure 10 Two lane single strip weigh pads installed on the MnRoad test site ri e bie Ki ge eevee gt tS ee Ni nant se e PS and v Figure 11 Comparison of a two lane left an
79. pstream sensor strip and the channel 1 Ch 1 signals come from the trailing downstream sensor strip with respect to the traffic direction It should be noted that the magnitude of the trailing sensor signals are bigger than that of the leading sensor signals This might be due to sensor characteristics i e the sensitivity of the trailing sensor strip expressed in terms of Coulombs Newton is higher Also notice that the signals from the trailing sensor strip have more ripples in the axle signals All piezoelectric sensors generate electricity not only from the direct load to the sensor but also from vibration It is evident from the rippling effect of the signals that the trailing sensor receives a high level of vibration energy During the test the researchers were able to visually observe a shockwave propagating from the leading to trailing edges of the dual strip weigh pad when truck wheels were moving from the leading to trailing sensor strips More specifically one can see multiple ripples or shockwaves in front of a turning wheel moving from the leading to trailing edges Energy transfer by this shockwave traverse would be negligible if the two sensor strips are separated by a sufficient distance since the vibration energy would be dampened before it reaches the trailing sensor strip In summary a pair of single strip weigh pads separated by a sufficient distance such as 14 ft would not be affected by this shockwave propagation while dual
80. r measurement is used an experimental study must be conducted to 59 determine useable life Since weigh pads must be replaced the next issue would be finding low cost replacement solutions Considering the high cost of BL sensors a new fabrication method in which the BL sensor is reused would lower the replacement cost This new fabrication method is recommended as a future study Piezoelectric materials generate charge signals proportionally to acceleration as well as to vibration WIM systems utilize the piezoelectric response to acceleration It is important to understand that piezoelectric sensors can generate charge signals in response vibration as well In particular large amplified charge signals are generated when the vibration matches with the sensor s resonance frequency In Section 4 2 amplified superfluous signals were observed during the experiments with a weigh pad that had an air cavity This signal was generated by propagation of vibration at the resonance frequency caused by air cavity in the sensor pad The superfluous signals disappeared when the air cavity was filled Therefore it is recommended that vibration damping material is used in the slot where the BL touches the re enforced rubber material Dampening the vibration force before it reaches the sensor strip would increase the accuracy of the WIM measurements During the installation wrinkles can be formed on the weigh pad as described in Section 4 1 and shown in
81. r truck which is a test truck available at the MnRoad facility was tested and one of the waveforms is shown in Figure 42 The known GVW of this vehicle was 79 720 pounds If the areas under the curve of the two test vehicle signals which would represent axle weighs are compared the factor is around 18 This is close to the weight ratio of the two vehicles This indicates that the axle waveforms approximately correspond to the axle loads of the two vehicles Accuracy tests are later discussed in Section 5 3 43 Volt 1 8 1 6 1 4 1 2 0 6 17 500 18 000 18 500 19 000 19 500 20 000 20 500 21 000 21 500 22 000 22 500 23 000 23 500 Sample Index Figure 41 Charge amp signals of a van Oct 16 2011 Volt 2 4 2 2 2 1 8 1 6 1 4 1 2 1 0 8 0 6 0 4 0 2 0 0 2 0 4 0 6 51 000 51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 55 500 56 000 56 500 57 000 Sample Index Figure 42 Charge amp signals of a five axle semi trailer truck Aug 16 2011 44 In all test cases the charge amps behaved as expected producing waveforms such as shown in Figures 41 and 42 close to the ideal axle model developed in Section 3 1 The resting level of the signals returned quickly to the signal ground when the load was removed i e the charges stored in the measurement capacitor were discharged with the designed time constant which is necessary in order to be ready for the next measurem
82. raw data using the Rec button The main usage of this real time plot is to quickly inspect the waveforms against visual observation of axles of a vehicle a frmGraphX lol x Channels Sensor Signals 0 500 1 000 1 500 2 000 2 500 3 000 3 500 4 000 yemin 0 5 ymax 5 5 _set Vv ch 0 V ch 1 Figure 30 Real time plot of ADC channels After recording the raw ADC data using the Rec button the user may need a plot tool to analyze the data For such a purpose a utility tool called WeighPad Plot was created as one of the tools available inside the weigh pad system This tool allows for the plot window to move forward or backward in the sample space as well as zoom in out using a setting of the Y range and or X range values A sample screen of the WeighPad Plot utility is shown in Figure 31 This is a useful tool to diagnose signal problems such as pad vibration problems or abnormal signal idle levels This utility is available as a separate program from the main weigh pad program 32 i i 2 1 8 1 6 v co 14 v cL v C2 v C3 1 2 1 16 000 17 000 18 000 19 000 20 000 21 000 22 000 23 000 24 000 25 000 F Research WIM WeighPAD WeighPad_TestData mnroad_data kK lt gt 20111016 143247 4096 4 bin Load File Y Range 1 2 XRange m Sekize 16000 gt 25500 Figure 31 Weigh Pad plot utility 3 2 2 Setting Wizards The weigh pad system require
83. s many system settings All of the user settings are listed under the Settings menu in the main window The items in this menu are Site Setup Axle Sensor Setup Calibration Factors Speed Adjustment Factors ESAL Setup Limit Parameters Signal Thresholds Selecting any of the menu items above would pop up a setup wizard that guides the user using easy to understand GUI entries Among the setups the Site Setup is explained here as an example The rest of the setup windows are summarized in Appendix C The Site Setup window is shown in Figure 32 All of the items in this window are prerequisite to system startup and must be set before any data collection is initiated The items include Site ID Location Lane setups classification definition and the data root directory The site ID should be a numeric number but the location can be any text that describes the location The site ID number is extremely important since all WIM data files produced are named using the date of the data collected and the site ID The lane direction and arrow direction combo box determines the lane direction and traffic flow direction displayed inside the vehShow controls The weigh pad software uses the identical vehicle classification algorithm and software components deployed in the BullConverter 21 which requires a class definition file The Browse button in the second GroupBox allows navigation of files and directories for the selection of a class def
84. s of a vehicle record For example when a calibration vehicle passes by the operator may freeze the vehShow control to carefully look at the details of the measured values A yellow Rec button can be seen next to the Pave Temp groupbox the third row from top of the window When this button is pressed the binary ADC outputs of the axle signals are recorded into a file A red Stop button appears right next to the Rec button which is used to stop the recording The recorded raw data can be later used for signal analysis The main window also includes a display of pavement temperatures either in Fahrenheit or Celsius depending on the user s selection A type k thermocouple must be taped on the pavement 31 and its lead must be plugged in to the console box to properly display the temperature otherwise N A is displayed Under the graph menu there is a menu item called Show Graph When this item is selected real time axle signals from a pair of upstream and downstream channels assigned to a lane are plotted A sample screen is shown in Figure 30 Because the graph must be displayed while computing the axle weights and all of the WIM values only 256 samples per second out of 4 096 samples are displayed Thus the plot appears blocky It displays one lane at a time and the display lane can be selected using the Channels menu If more detailed signal analysis is necessary the operator should always record the
85. s path can be designed to only remove slowly changing charge signals and it is often called a DC servo loop In the actual circuit a discharge path with a 20 second time constant was added This means that any charge signals that do not change for the duration longer than 20 seconds are gradually dissipated to zero In effect it removes the DC component of the charge signals that are generated by heat or other factors The time constant of the axle signals in the charge amp is set at a half second so that any signal that has a rate of change less than a half second is passed through without activating the dissipation path The basic charge amp circuit is shown in Figure 13 A typical charge amp with T resistor network is shown in the first stage consisting of Ul R2 R3 R4 and C1 The resistor RI is used to protect the input stage of Ul a 500 Ohm resistor was used The capacitor value of C1 determines amplification and the Vout signal range Since the sensitivity of BL sensors is around 49 pC N a high Newton value corresponding to a single signal component of an axle load is about 34 000 Newton The charge signal generated by 34 000 Newton is 34 000 x 49 1 666 000 pC Since V Q C if 5V is used as the peak voltage the C1 value comes out to be 0 33 uF However this value has to be reduced by a factor influenced by the DC servo loop that pulls down the overall signals as described in the previous paragraph The final C1 value selected was 0 06 uF
86. s window a9 Parameter Limits O x Maximum vehicle length possible 75 feet Maximum axle spacing possible 45 feet Maximum tire footprint length possible Rp feet Sensor width 7 millimeters Minimum possible speed on this road is mph Maximum possible speed on this road f 10 mph Appendix C Sample Weigh Pad WIM Data veh Lane Time Axle speed AS1 feet AS2 AS3 AS4 AS5 AS6 AS7 AS8 AS9 AS10 AS11 AW1 kips AW2 AW3 AW4 AW5 AW6 AW7 AW8 AW9 AW10 AW11 AW12 GVW Class Err 100thSec pavTemp 1 1 11 21 57 2 69 8 3 3 13 2 93 6 06 2 0 70 37 6 2 1 11 22 00 3 69 12 0 13 3 4 74 4 03 2 27 11 04 3 0 55 37 5 3 1 11 22 27 3 66 18 7 4 6 12 10 7 94 14 23 34 27 6 0 41 36 4 4 2 11 22 28 2 69 11 8 5 13 3 32 8 45 3 0 59 36 4 5 2 11 22 30 2 76 14 1 4 63 5 33 9 96 5 0 98 36 2 6 1 11 22 50 2 65 9 5 3 42 3 46 6 88 2 0 4 35 7 7 1 11 23 53 2 59 11 9 4 85 5 08 9 94 3 0 45 35 5 8 1 11 24 11 0 61 55559990707977007720 00 15 8 0 35 7 9 1 11 24 35 2 72 11 5 4 88 3 05 7 92 3 0 59 35 7 10 1 11 24 40 0 67 555191909799000017770 00 15 8 0 35 9 11 2 11 24 50 2 70 8 6 2 13 1 47 3 60 2 0 51 36 0 12 1 11 24 55 2 67 9 7 3 26 3 02 6 27 2 0 98 36 1 13 2 11 24 57 0 20 55997979790000017770 00 15 8 0 36 1 14 1 11 25 30 5 67 16 6 4 3 36 1 4 0 16 41 23 41 18 15 13 2
87. search team supplied the RoadTrax BL sensors described in Section 3 4 to IRS and the IRS technician cut long grooves for the BL sensor strips and coaxial cable glued two pads together along with the BL sensor strips and the lead cables and then sanded edges to produce smooth gradual leading and trailing edges Air cavity was created accidently by leaving a portion of groove unfilled The location of unfilled groove is illustrated in Figure 37 IRS technicians created two parallel grooves for the entire 24 feet 7 32 m of the pad although the Channel 0 CO groove is only needed for the first half 12 ft The Channel 3 C3 groove of the Weigh Pad 1 was filled by a 12 ft 3 6 m BL sensor and the lead coaxial cable along with adhesives so no air cavity exists in C3 However only the half of the Channel 0 groove of the Weigh Pad 1 was filled by a 12 ft 3 6 m BL sensor and the remaining groove was left empty forming an air cavity The second weigh pad produced Weigh Pad 2 did not have this problem because IRS technicians did not make unfilled grooves as the Weigh Pad 1 in Figure 37 39 In order to test these two different weigh pads two 24 feet sensor pads were installed in the MnRoad low volume road The test vehicle and installed weigh pads are shown in Figure 38 The test vehicle traveled in the direction shown by the large arrow in Figure 37 When the vehicle travels in this direction the signals in CO and C1 should be close to the sign
88. strip weigh pads do The dual strip weigh pads are less accurate while they use more material and are bigger in size and heavy in weight The research team concludes that the idea of dual strip weigh pads is not a good one for a portable WIM system 11 3 2 8 2 6 2 4 2 2 2 1 8 vw vi Cho v Chi Volt o 0 2 0 4 0 6 0 8 1 41 000 41 500 42 000 42 500 43 000 43 500 44 000 44 500 45 000 45 500 46 000 Sample Index Figure 7 Five axle semi trailer truck WIM signal recorded from a single lane dual strip weigh pad 2 2 3 Two Lane Weigh Pad Design Most of rural roads targeted roads of weigh pads are two lane roads It is thus more practical to develop a pair of two lane weigh pads than two pairs of single lane weigh pads After a successful demonstration of single lane prototypes the MnDOT TDA Office of Transportation Data amp Analysis recommended the research team to develop two lane weigh pad prototypes as an additional task Learning from the single lane weigh pad experiences the clear choice of the two lane design is creating two lane single strip weigh pads One of the difficulties in creating two lane length is that the coaxial cable of the sensor strip from the far lane must run through the weigh pad material without affecting the near lane sensor strip running in parallel This problem was solved by milling out the coaxial cable and sensor strip slots in parallel as well as increasing the weigh
89. t i e charges are not only produced by loads but also by heat In order to overcome this problem a charge amp integrated with a heat effect filtering circuit was developed and integrated Lastly the system must run on a battery and must be able to sustain its operation for a certain period of time Due to the heavy computations involved in WIM systems it is not easy to design a computer system that uses only a small amount of energy and yet provides a sufficiently high computational performance To strike the balance of the system cost performance and battery time the current system is designed to continuously run for minimum 24 hours with the built in internal battery pack External battery packs must be used if the system has to run longer than 24 hours The proposed and envisioned portable WIM system in this project was successfully developed A complete working prototype was built and tested This report describes all aspects of the developed weigh pad system Chapter 2 describes the hardware designs which include weigh pad charge amp and the console computer designs Chapter 3 describes the software part of the system including the axle computational model and the overall system software Chapter 4 shows a highway installation method and discusses issues related to air cavity in the sensor pad The weigh pads were initially tested on the MnRoad facility and then later tested on real highway traffic Chapter 5 summarizes various test r
90. t a time Removing the sensor pads took a total of 14 minutes or 7 minutes per lane Table 6 summarizes installation and removal time of weigh pads The weigh pads did not show any signs of damage after running 24 hours of run under the TH 53 traffic that included many five axle semi trailer trucks The date of this installation was November 4 and the pavement was covered with frost in the early mornings Although the research team expected that this cold temperature may stiffen the sensor pads and result damages but no such evidence was found Both the MnRoad and the TH 53 tests were performed on bituminous pavements Most lower ADT roads in Minnesota have bituminous surface so this technology was note concrete pavements Figure 35 Sleeve anchor screws are fastened in 2 ft spacing 38 Figure 36 Some portions had wrinkles that caused vibration and error on the axle signal Table 6 Weigh Pad Installation and Removal Time Installation Removal Single Lane 15 minutes 7 minutes two 12 ft weigh pads Two Lanes 30 minutes 14 minutes two 24 ft weigh pads 4 2 Air Cavity and Vibration Problems During the initial prototype weigh pad tests it was accidently discovered that existence of air cavity in the sensor pads adversely affects the axle load signals The sensor pads were fabricated by a local company called the Industrial Rubber and Supply IRS which specializes in customized conveyor belts The re
91. tallation planning Figure 34 Weigh pad installation at Cotton Minnesota TH 53 Figure 35 shows a segment of the completed installation of a sensor pad The spacing between two sleeve anchor screws shown is two feet The taped edges can be seen in the picture Although it cannot be seen from Figure 35 some part of the pads was not tightly fastened to the pavement Figure 36 shows a defect portion of the installation This picture was taken after one day of operation and it was clear that the pads were not stretched and tightened before placing the anchor screws consequently wrinkles were formed in few places Initially the research team 37 assumed that taping would hold down the wrinkled portion of the pad on the pavement It turns out that taping alone was not sufficient to hold down the wrinkled portion of the pad The pad edges were ripped out of the tape after passing of just a few vehicles creating a space between the pavement and weigh pad as shown in Figure 36 According to the recorded sensor signals and data these wrinkles in the pad caused vibration and created false axle signals reducing the accuracy of measurements The lesson learned from this installation experience is that the pads should be laid flat and stretched before fastening the sleeve anchors The next day the sensor pads were removed and the removal time was measured For removal a MnDOT traffic control crew was again called in and the lanes were closed one a
92. tes the real data points Table 11 summarizes the computed correlation coefficients and the coefficients of determination between the IRD and weigh pad data Notice that every coefficient is above 0 9 which indicates a very strong correlation Among these coefficients length i e axle spacing coefficients are above 0 99 which indicates both data are nearly identical GVW and speed also have strong linear relationships correlation coefficients exceeding 0 96 Table 11 Correlation Coefficients and R Between IRD and Weigh Pad data Measurements Correlation Coefficient R GVW 0 965023 0 9313 Speed 0 966612 0 9343 Length 0 999721 0 9994 In the above fit tests vehicle length had the best linearity relationship and the best goodness of fit GVW and speed had strong linear relationships but the percentage of linearly related data drops to 93 percent Notice from the scatter graphs in Figures 48 50 that vehicle length axle spacing and speed had very tight linear relationships with almost no exceptions On the other hand GVW was not tightly bunched to the linear line Again these plots confirm that weight data had the least consistency which agrees with Table 11 53 Weigh pad GVW Kips Weig pad Speed mph 100 90 80 70 60 50 40 30 20 10 IRD GVW Kips Figure 48 Scatter plot of IRD vs Weigh Pad GVW data IRD Speed mph Figure 49 Scatter plot of IRD vs Weigh Pad speed data 5
93. tforthisproject 17 Figure 15 Measurement Computing USB 7202 200K S s 16 bit ADC 18 Figure 16 Access I O Products Inc USB AI16 16A 500K S s 16 bitADC 18 Figure 17 NI USB 62 102 200K S s 16 bit ALDM anaya Gla nana Gele 19 Figure 18 PCI DAS6013 ADC boande i i etilen iadeye inedsnna da a dayana S re iy basil aa 19 Figure 19 Console computer block lap dim 14 ai li le das ee iye 21 Figure 20 Console computer enclosure specification 22 Figure 21 Custom enclosure built using aluminum sheets 23 Figure22 Middlelayeroftheconsolecomputer 23 Figure 23 Top layer of the console Computers a5 io yal Balda ae alel ala ali 24 Figure24 Backsideoftheconsolecomputerenclosure 24 Figure 25 Hardware in Loop HIL WIM signalsimulator 25 Figure 26 Gaussian axle signal model x miele senlik ni a i ia e sa e dveatvenseuns 26 Figure 27 Plot of the Example 1 axle 008 dicen a ala al 28 Figure 28 Net component vehShow dll developed for visual modeling of individual vehicle records yalan ya an lm laleli a l il al ml dalan ma alimli elmali 30 Figure 29 GUI screen shot ofthedevelopedweigh padconsole 31 Figure 30 Real time plot of ADC channel5
94. the initial settings It comes down to fact that there is less freedom and more limitation in designing a system with a laptop computer Consequently the research team decided to use a mother board and build all necessary components from the mother board The motherboard selected for the computing unit belongs to a form factor called mini ITX and they are commonly used in embedded PC applications The processor used is an Intel 1 66GHz dual core Atom which consumes less energy while it provides sufficient computing power through dual cores Another important aspect of mother boards is the availability of a PCI slot that can interface with a PCI based ADC board In general a PCI based ADC board is more reliable and provides a better data transfer bandwidth than the USB based boards The mother board also includes an external Video Graphics Array VGA port which can connect an open frame LCD monitor with a hard On OFF power switch instead of just a sleep state in a laptop PC case The specification of the motherboard used for the weigh pad console is summarized in Table 3 Table 3 Console Mother Board Processor 1 66GHz Dual Core Atom D510 Memory 240 pin DDR2 800 DIMM 2GB Hard Disk 2 5 Segate SATA 5400rpm 160GB OS Windows XP Pro Embedded Keyboard Mouse PS2 Display Port VGA Form Factor Mini ITX Battery External not included VO Connectors USB PCI RS 232 COM port LAN 10 100 1000 M bps Ethernet 20
95. was below 1 This test indicates that the speed effect is less significant than the tests observed from the MnRoad tests We believe that the differences in setup resulted in a different effect For the MnRoad speed tests the sensor pads were fastened to the pavement only using tapes For the TH 53 tests the sensor pads were fastened using sleeve anchors with washers and then the edges were taped This installation difference appears the cause of the differences in speed effect 55 In summary speed impact appears less significant according to Table 12 as the sensor pads are more tightly fastened to the pavement but oscillation of physical weights by the suspension system of vehicles seems a more dominant factor in the GVW differences between the two systems It also suggests the accuracy of WIM measurements is limited by the suspension oscillation effect and cannot be improved by instrumentation accuracy of any WIM systems Table 12 m GVW Ratio over GVW Ranges in Kips and Speed Ranges in mph GVW Range in Kips mm e ee 0 0 0 0 0 0 0 0 0 0 1 34 1 19 56 Table 13 Number of Vehicle Records in the Defined Range GVW Range in Kips m aaa a 12 1 2 1 2 2 2 0 0 1138 The last test is to compare vehicle classification results of the two systems using the same class definition table According to the GVW R about 7 percent of the weight data did not follow linearity This test is to investigate whether this
96. weight and the corresponding calibration factor calfac is defined by calibrated_weight calfac recorded_weight 12 This linear regression estimate however exhibits poor fit in the region of speeds less than 20 mph 32 Km h and also in the above 70 mph 112 7 Km h Therefore a better fit function is desirable 47 weight Kips 0 10 20 30 40 50 60 70 80 90 speed mph Figure 45 Scatter plot of speed vs weight of the same vehicle and linear regression Table 7 Linear Calibration Factors multiplication factors for Different Speeds 5 mph 65 mph 85 mph calfac 1 3089 1 1366 1 0043 0 8996 0 8147 0 7444 0 6853 0 6349 0 5914 A better fit function chosen is a log fit function This fit function is shown in Figure 46 Note that it has a better fit in the areas of below 25 mph and above 70 mph speeds This log fit function is given by y 2 1248In x 2 2045 13 where x is the speed in mph and y is the vehicle weight in Kips This function has R 0 8566 which tells us that the log function is a better fit than the linear regression R 0 7897 Substituting x with the computing speed and dividing this by 4 6 Kips static weight of the test vehicle gives the inverse of calfac multiplication factor The final calfac computed for different speeds using the log estimate is summarized in Table 8 Because it is a log function the curve drops too quickly in the area of below 10 mph Therefore
97. wo 100 feet coaxial cables The advantage of embedding dual sensor strips in a single pad would be the known distance between two sensors set at the factory level It simplifies the installation process by eliminating the user responsibility of measuring the sensor spacing by installing only one pad However there were three critical drawbacks observed during the initial tests The first was the weight of the weigh pad The weight of the Figure 5 dual strip weigh pad was measured at 55 Ib 23 Kg and it was heavy for carrying around Second since it requires much more pad material the cost of the sensor pad was significantly increased Third the tail portion of the axle signals were bouncy due to a shockwave propagation which will be explained later CE Figure 4 Dimensions of single lane dual strip weigh pad Figure 5 A single lane dual strip weigh pad prototype In order to test the charge signals of the constructed weigh pads the sensor outputs were first connected to charge amps and then the output signals of the charge amp were observed using an oscilloscope The weigh pads responded very well for human weights when it was simply tested by stepping onto it Next the research team tested the weigh pads using test vehicles This time the charge amp outputs were connected to an ADC board in order to save the waveform The first vehicle test was performed at the low volume road of the MnRoad facility on June 4 20

Download Pdf Manuals

Related Search

Related Contents

  Brocade DCFM v10.3.0 Release Notes v1.0  Camera Manual - EOC  Digitus DC-40200 video converter  metodologia de fiscalização de obras  USER MANUAL - AvProSupply  MergeCOM-3™ C/C++ DICOM Toolkit Message  施工マニュアル  Catalogo completo    

Copyright © All rights reserved. Failed to retrieve file