Development of a Weigh-Pad-Based Portable Weigh-In

Development of a
Weigh-Pad-Based Portable
Weigh-In-Motion System
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 request 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/RC 2012-38
4. Title and Subtitle
5. Report Date
Development of a Weigh-Pad-Based Portable Weigh-In-Motion
System
December 2012
7. Author(s)
8. Performing Organization Report No.
6.
Taek M. Kwon
9. Performing Organization Name and Address
10. Project/Task/Work Unit No.
Department of Electrical and Computer Engineering
University of Minnesota Duluth
1023 University Drive
Duluth, MN 55812
11. Contract (C) or Grant (G) No.
12. Sponsoring Organization Name and Address
13. Type of Report and Period Covered
Minnesota Department of Transportation
Research Services
395 John Ireland Boulevard, MS 330
St. Paul, MN 55155
Final Report
CTS Project #2009020
(C) 89261 (WO) 114
14. Sponsoring Agency Code
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 sideby-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, Weigh in motion
scales, Weighing devices
No restrictions. Document available from:
National Technical Information Services,
Alexandria, Virginia 22312
19. Security Class (this report)
21. No. of Pages
Unclassified
20. Security Class (this page)
Unclassified
87
22. Price
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 and 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
Chapter 1:
Introduction ........................................................................................................... 1
Chapter 2:
Hardware Design .................................................................................................. 5
2.1 Overall System ...................................................................................................................... 5
2.2 Axle Weight Sensors and Weigh-Pad Design ...................................................................... 5
2.2.1 Axle Sensor and the Pad Material ................................................................................. 5
2.2.2 Single-Lane Weigh-Pads Design ................................................................................... 7
2.2.3 Two-Lane Weigh-Pad Design ...................................................................................... 12
2.3 Charge Amplifier ................................................................................................................ 15
2.4 Analog-to-Digital Converter (ADC) ................................................................................... 17
2.5 Console Box and Enclosure ................................................................................................ 20
Chapter 3:
Weigh-Pad System Software Design ................................................................. 25
3.1 Axle Computational Model ................................................................................................ 25
3.1.1 Signal Modeling and Digital Signal Generation ......................................................... 26
3.1.2 Computing the Digitized Signals Back to Weight for Verification .............................. 28
3.2 System Software ................................................................................................................. 29
3.2.1 Overall GUI and Tool Sets .......................................................................................... 29
3.2.2 Setting Wizards ............................................................................................................ 33
3.2.3 Output Data Format .................................................................................................... 34
Chapter 4:
Weigh-Pad Pavement Installation ..................................................................... 36
4.1 Installation on Highways .................................................................................................... 36
4.2 Air Cavity and Vibration Problems .................................................................................... 39
Chapter 5:
Experimental Results .......................................................................................... 43
5.1 Charge Amp Tests .............................................................................................................. 43
5.2 Experiments on Influence of Speeds on Weight ................................................................. 45
5.2.1 Theory .......................................................................................................................... 45
5.2.2 Data Collection ............................................................................................................ 46
5.2.3 Analysis ........................................................................................................................ 47
5.3 Side-by-Side Tests of Weigh-Pad Vs. IRD WIM Systems ................................................ 49
5.3.1 Test Setup and Data Collection ................................................................................... 49
5.3.2 Data Analysis ............................................................................................................... 51
Chapter 6:
Conclusions and Future Recommendations ..................................................... 59
6.1 Conclusions ......................................................................................................................... 59
6.2 Future Recommendations ................................................................................................... 59
References
............................................................................................................................... 61
Appendix A: Weigh-Pad Test Pictures
Appendix B: Weigh-Pad System Setting Wizards
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, single-strip weigh-pad ......................................................... 8
Figure 3: A single-lane, single-strip weigh-pad prototype ............................................................. 8
Figure 4: Dimensions of single-lane, dual-strip weigh-pad ............................................................ 9
Figure 5: A single-lane, dual-strip weigh-pad prototype .............................................................. 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 weighpad ................................................................................................................................................. 12
Figure 8: Dimensions of the Two-lane, single-strip weigh-pad: top view ................................... 13
Figure 9: Prototypes of a pair of two-lane, single-strip weigh-pads ............................................. 13
Figure 10: Two-lane single-strip weigh-pads installed on the MnRoad test site.......................... 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
weigh-pads .................................................................................................................................... 15
Figure 13: Basic charge amp circuit ............................................................................................. 17
Figure 14: A prototype two-channel charge amp built for this project......................................... 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-bit ADC ............................. 18
Figure 17: NI USB-6210: 200K S/s, 16-bit ADC ......................................................................... 19
Figure 18: PCI-DAS6013 ADC board .......................................................................................... 19
Figure 19: Console computer block diagram ................................................................................ 21
Figure 20: Console computer enclosure specification .................................................................. 22
Figure 21: Custom enclosure built using aluminum sheets .......................................................... 23
Figure 22: Middle layer of the console computer ......................................................................... 23
Figure 23: Top layer of the console computer .............................................................................. 24
Figure 24: Back side of the console computer enclosure ............................................................. 24
Figure 25: Hardware-in-Loop (HIL) WIM signal simulator ........................................................ 25
Figure 26: Gaussian axle signal model ......................................................................................... 26
Figure 27: Plot of the Example-1 axle signal................................................................................ 28
Figure 28: .Net component “vehShow.dll” developed for visual modeling of individual vehicle
records ........................................................................................................................................... 30
Figure 29: GUI screen shot of the developed weigh-pad console ................................................ 31
Figure 30: Real time plot of ADC channels.................................................................................. 32
Figure 31: Weigh-Pad plot utility ................................................................................................. 33
Figure 32: Site setup window ....................................................................................................... 34
Figure 33: Tools needed for WPad installation ............................................................................ 36
Figure 34: Weigh-pad installation at Cotton, Minnesota, TH-53 ................................................. 37
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 ......... 39
Figure 37: Location of weigh-pad air cavity................................................................................. 41
Figure 38: Installed weigh-pads with air cavity and the test vehicle ............................................ 41
Figure 39: Air cavity generated noise on Channel-0 (C0). ........................................................... 42
Figure 40: Removing Channel-0 signal clears the superfluous signals ........................................ 42
Figure 41: Charge amp signals of a van (Oct 16, 2011) ............................................................... 44
Figure 42: Charge amp signals of a five-axle semi-trailer truck (Aug 16, 2011) ......................... 44
Figure 43: Force and slope relationship ........................................................................................ 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 .............................................. 47
Figure 45: Scatter plot of speed vs. weight of the same vehicle and linear regression. ............... 48
Figure 46: Log regression of weight data by different speeds ...................................................... 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
Figure 49: Scatter plot of IRD vs. Weigh-Pad speed data ............................................................ 54
Figure 50: Scatter plot of IRD vs. Weigh-Pad vehicle length data ............................................... 55
Figure 51: Classification comparison between the IRD and Weigh-Pad system vehicle records 58
List of Tables
Table 1: RoadTrax BL Sensor Specifications ................................................................................. 6
Table 2: Weigh-Pad Material Specifications .................................................................................. 7
Table 3: Console Mother Board .................................................................................................... 20
Table 4: Key Console Box Components....................................................................................... 21
Table 5: Weigh-Pad CSV Column Format ................................................................................... 35
Table 6: Weigh-Pad Installation and Removal Time .................................................................... 39
Table 7: Linear Calibration Factors (multiplication factors) for Different Speeds ...................... 48
Table 8: Log Calibration Factors .................................................................................................. 49
Table 9: Setup Parameters ............................................................................................................ 51
Table 10: Limit Parameters ........................................................................................................... 51
Table 11: Correlation Coefficients and R2 Between IRD and Weigh-Pad data ............................ 53
Table 12: Average GVW Ratio over GVW Ranges in Kips and Speed Ranges in mph .............. 56
Table 13: Number of Vehicle Records in the Defined Range ...................................................... 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, 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 inpavement 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 of
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 inpavement 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 ºF, and the corresponding gross vehicle
weights (GVWs) translated from the axle waveforms 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.
The 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 R2, 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 inpavement 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
comparison 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 $220,000 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 portable
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
1
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 the 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&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 axle 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&D efforts 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, long, 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 heat, 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 weighpad, 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.
3
The weigh-pads were initially tested on the MnRoad facility and then later tested on real
highway traffic. Chapter 5 summarizes various test results and analysis. Chapter 6 concludes the
report with final remarks and future recommendations.
4
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
Weight
Sensor
A/D
Converter
Charge
Amplifier
Computing
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 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).
5
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 ≤ C ≤ 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 in 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 1 mm
to reduce the bumpiness.
6
Table 2: Weigh-Pad Material Specifications
Product Model
Transtex Product Code
Product Construction
UTILITY 2-160 GRADE II 1/32X1/32-NA
908860
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 lbs/ft2(Kg/m2)
0.98 ± 10%
Rated Working Tension
160 lbs/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 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 lbs (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.
7
6"
144"
Figure 2: Dimensions of single-lane, single-strip weigh-pad
Figure 3: A single-lane, single-strip weigh-pad prototype
8
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 two 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 lb
(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.
4"
24"
32"
4"
144"'
Figure 4: Dimensions of single-lane, dual-strip weigh-pad
9
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, 2010.
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
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 singlelane, dual-strip weigh-pad, are shown in Figure 7. In the graph, the channel-0 (Ch-0) signals
come from the leading (upstream) 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-strip weigh-pads do. The dualstrip 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
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 & 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 weighpad 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 singlelane 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 (C0) 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
12"
6"
1.5"
4.5"
3"
12"
1"
9"
3"
9"
3"
4.5"
1.5"
Sensor Strips
Coax Cables
Figure 8: Dimensions of the Two-lane, single-strip weigh-pad: top view
Figure 9: Prototypes of a pair of two-lane, single-strip weigh-pads
13
Figure 10: Two-lane single-strip weigh-pads installed on the MnRoad test site
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 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 ⁰F (71.1 ⁰C) 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 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 200⁰F, 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. This 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 U1, R2, R3, R4, and C1. The resistor R1 is used
to protect the input stage of U1; 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
µF. 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 µF. 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*R5), 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, R5=R6=10M Ohm and C2=C3=2.2 µF, 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
C2
+Vs
R6
−
R4
U2
+
R3
R2
-Vs
C1
R1
R5
C3
+Vs
−
Vout
U1
WIM
Sensor
+
-Vs
Figure 13: Basic charge amp circuit
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 choice. 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, USBAI16-16A (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
All 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 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 in 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 equipped 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 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 openframe 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
I/O Connectors
USB, PCI, RS-232 COM port
LAN
10/100/1000 M bps Ethernet
20
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
SVGA (800 x
600) LED back
light LCD
16bit 200K S/s
ADC
LCD SW
PCI
Slot
Type-K USBbased
Thermocouple
USB
VGA
Atom-Based Mini-ITX
12V
Keyboard +
Mouse
USB
DC
Power
Master SW
Li-Ion Polymer
Battery Bank
(14.8V)
Li-Ion Polymer
Charger
AC
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 connectors. 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.
17"
11"
2 1/2"
3 1/4"
2 3/4"
3"
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:
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 the 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 weighpad 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
research, 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.
f ( x) = ae − ( x −b )
∫
+∞
−∞
2
/ c2
(1)
f ( x) = ac π
(2)
In Eq. (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
a
b
5c
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 pound
Sampling Rate: R Samples/Sec
Footprint Length: l feet
Speed: v foot/sec
Sensor width: w feet
Multiplication factor: f pound
Since the signal support area is the footprint of the tire and sensor width, it can be expressed as:
5c =
l+w
sec
v
Therefore, c is computed as:
c=
l+w
5c
(3)
The time it takes to go through a single sensor width is,
tw =
w
v
sec
(4)
Let the area under the axle signal be Asig , i.e.
26
+∞
f ( x)dx ac π
∫=
Asig
=
(5)
−∞
Deriving from Eq (1), if the signal is continuous, the signal area and weight are related as:
W pound
=
Asig
tw
× f pound
Thus,
Asig =
W pound tw
(6)
f pound
Equating Eqs. (5) and (6) gives,
ac π =
W pound tw
f pound
Finally, a is computed as:
a=
W pound tw
(7)
f pound c π
Collecting these results, the final axle waveform with computed parameters of a and c is
obtained as:
2
2
f ( x) = ae − ( x −b ) / c
where a and c are give as:
W pound tw
a=
f pound c π
c=
for 0 ≤ x ≤ 1
l+w
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: W pound = 20,000 pounds
Sampling Rate: R = 4000 S/sec
Footprint Length: l = 0.656167979 feet
Speed: v = 88 foot/sec
Sensor width: w = 0.16404199475 feet
Multiplication factor: f pound =(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 ( xi ) are generated as:
For i=0 to 3999
{
//shift of signal to right 2000 points
xi= (i − 2000) / 4000
2
f ( xi ) = ae − xi / c
2
}
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 < i < 2000 + (2.5c)4000
Figure 27: Plot of the Example-1 axle signal
3.1.2 Computing the 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
Asig = ∑ vi
(8)
i
=
W pound
where
Asig f pound Asig f pound
 f pound 
= = Asig 
v
tw R
( w / v) R
 wR 
(9)
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.,
Asig = 96.74431
tw =
0.164
sec
88
W pound
96.74431× (1541.4765 × 2)
= 20, 000 pounds
tw × 4000
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 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 (GUI) 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.
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
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 details 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 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.
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 Xrange 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
Figure 31: Weigh-Pad plot utility
3.2.2 Setting Wizards
The weigh-pad system requires 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 definition 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.
Figure 32: Site setup window
3.2.3 Output Data Format
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
Number
1
Column name
2
3
4
5
6-16
29
Lane#
Time
AxleC
Speed
Axle Spacing (AS):
AS1,…,AS11
Axle weights(AW):
AW1,…,AW12
GVW
30
Class
31
32
Err#
100thSec
Vehicle record is numerically indexed and this
column shows the index
Lane number of the vehicle passed through
Time in hh:mm:ss where hh is military hour
Number of axles on the vehicle
Speed of the vehicle in mph
Axle spacing in feet. It contains total 11 fields
separated by comma.
Axle weights in Kips. It contains total 12 fields
separated by comma.
Gross Vehicle Weight (GVW) in Kips. It is simply a
summation of each axle weight.
Vehicle class determined by the classification
algorithm
It is a numeric code that represent an error
100th seconds of the time in Column 3.
33
pavTemp
Pavement temperature in Fahrenheit
17-28
Index
Description
35
Chapter 4:
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 weighpads 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 flatwasher in the middle of the weigh-pad. All four approaches were tested at the 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 & 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 ¼” diameter , 6” long
drill bits
Strong bonding utility table - black color
(black Gorilla tape was used)
Sleeve Anchor, ¼ “ diameter, 2-1/4”
length with ¼” 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
¼” using a hammer drill, and then a sleeve anchor is inserted to the drilled hole after placing a
¼” washer. The screw on top of the sleeve anchor is turned clockwise to expand the sleeve,
which fastens the washer and thus the pad to pavement. Sleeve anchors are installed by spacing
approximately 2 feet apart, requiring 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, doublesided (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 installation 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 at 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 4th 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 research 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 (C0) 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 C0 and C1 should be close to the signal 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 C0
has signals that appear for every axle signals of C2 and C3 even though no axle load was present.
To show that C0 signals are false, plot of C0 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 C0 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 C0 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 C0, 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 C0 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 passes through a weigh-pad, the weigh-pad acts as a small bump that creates a vibration.
This vibration is propagated to C0 but it is amplified by the air cavity before it reaches to the C0
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 C0 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
Air Cavity
Weigh-Pad 1
Ch0
Ch3
14'3"
RoadTrax
BL Sensors
13'9"
Test Vehicle
Travel Direction
Ch2
Ch1
Weigh-Pad 2
Yellow Lane
Markings
Figure 37: Location of weigh-pad air cavity
Figure 38: Installed weigh-pads with air cavity and the test vehicle
41
Figure 39: Air cavity generated noise on Channel-0 (C0).
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 350ºF 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 remove 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-trailer 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
Figure 41: Charge amp signals of a van (Oct 16, 2011)
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 measurement.
To examine the pavement heat effect, the weigh-pads and charge amps were tested under the
pavement temperature range of 85 – 134 ºF. 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 ºF
using a heat gun. The effects to the axle waveforms by the tested temperature range (85 – 134 ºF)
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 even 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 θ 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 Fh sin(0°) =0 ; thus, the load received by the sensor, Fs,
is equal to the gravitational force only, i.e., the weight Fg. 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=
Fh + Fg
s
(10)
45
In theory, as the vehicle’s speed increases the horizontal force, Fh, applied to the load sensor
should increase while the gravitational weight of the vehicle remains 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.
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 ºF (19 - 24 ºC) 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, 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 highstrength 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 R² = 0.7897
(R² 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 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
9
8
7
weight, Kips
6
5
y = 0.0533x + 3.2478
R² = 0.7897
4
3
2
1
0
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
speed
calfac
5 mph
1.3089
15 mph 25 mph 35 mph 45 mph 55 mph 65 mph 75 mph 85 mph
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:
(13)
y = 2.1248ln(x) - 2.2045
where x is the speed in mph and y is the vehicle weight in Kips. This function has: R² = 0.8566,
which tells us that the log function is a better fit than the linear regression (R² = 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, 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 nonlinear 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 effect would not diminish. Collection of more data is
recommended to finalize the speed effects.
Table 8: Log Calibration Factors
speed
calfac
5 mph
1.7692
15 mph 25 mph 35 mph 45 mph 55 mph 65 mph 75 mph 85 mph
1.2959 0.9925 0.8598 0.7818 0.7290 0.6902 0.6600 0.6358
9
8
7
weight, Kips
6
5
y = 2.1248ln(x) - 2.2045
R² = 0.8566
4
3
2
1
0
0
10
20
30
40
50
60
70
80
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. 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 & 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 TH53 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 concern for physical sensor-pad damages, but 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 #
Sensor Strip Spacing Lane 1
Lane 2
Sensitivity
Lane 1
Calibration Factor
Value
14 ft
14 ft
Upstream=6.25
Downstream=6.25
Lane 2 Upstream=6.25
Downstream=6.25
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
45 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 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 weighpad 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 as a 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 subsequent analysis, the weighpad 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 Square Error (RMSE) is used as the first similarity test. The RMSE is defined by:
N
RMSE =
∑ (x − y )
i =1
i
2
i
(14)
N
where xi is the IRD data and yi is the weigh-pad data. 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:
•
•
•
GVW RMSE = 4.10696 Kips
Speed RMSE = 1.28684 mph
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,
i.e.,
NRMSE =
RMSE
ymax − ymin
(15)
The NRMSEs computed according to Eq. (15) are:
•
•
•
GVW NRMSE=0.03880 (3.88%)
Speed NRMSE=0.02219 (2.219%)
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 of the sensor quality. 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 oneto-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
Corr ( X , Y ) =
Cov( X , Y )
σ xσ y
(16)
where Cov( X , Y ) is the covariance between two random variables X and Y, and σ x and σ y 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 > 0.8 suggests a strong linear relationship [20].
The coefficient of determination, denoted as R2, is also computed [20]. This value represents a
statistical measure of how well the regression line approximates 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 R2 Between IRD and Weigh-Pad data
Measurements Correlation Coefficient
R2
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
160
Weigh-pad GVW (Kips)
140
120
100
80
y = 1.1885x + 0.465
R² = 0.9313
60
40
20
0
0
20
40
60
80
100
120
IRD GVW (Kips)
Figure 48: Scatter plot of IRD vs. Weigh-Pad GVW data
100
90
Weig-pad Speed (mph)
80
70
60
y = 0.9656x + 3.0163
R² = 0.9343
50
40
30
20
10
0
0
20
40
60
80
IRD Speed (mph)
Figure 49: Scatter plot of IRD vs. Weigh-Pad speed data
54
100
80
Weigh-pad Veh Length (feet)
70
60
50
y = 1.0043x + 0.0255
R² = 0.9994
40
30
20
10
0
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 weighpad 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 = GVWWPad/GVWIRD
(17)
Here, GVWWPad is the raw weigh-pad GVW 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 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: Average GVW Ratio over GVW Ranges in Kips and Speed Ranges in mph
speed
0->20 mph
20->30
mph
30->40
mph
40->50
mph
50->60
mph
60->70
mph
80->90
mph
90-> mph
GVW Range in Kips
0102030405060708090>10
>20
>30
>40
>50
>60
>70
>80
>90
>
0
0
0
0
0
0
0
0
0
0
1.16
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1.61
2.14
0
0
0
0
0
0
0
0
1.34
1.4
1.3
1.57
1.62
1.33
1.13
1.1
0
0
1.36
1.23
1.45
1.43
1.37
1.4
1.18
1.16
1.04
1.19
1.09
1.31
1.34
1.26
1.01
1.53
1.18
0.93
1.3
5
0
0.95
0.37
0
0
0
0
0
0
0
0
56
Table 13: Number of Vehicle Records in the Defined Range
GVW Range in Kips
speed
0->20 mph
20->30
mph
30->40
mph
40->50
mph
50->60
mph
60->70
mph
80->90
mph
90-> mph
0>10
10>20
20>30
30>40
40>50
50>60
60>70
70>80
80>90
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
90>
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
0
0
0
71
12
1
2
1
2
2
2
0
0
1506
139
19
66
29
15
20
61
16
1
1138
70
2
9
5
1
2
16
5
0
18
1
0
0
0
0
0
0
0
0
The last test is to compare vehicle classification results of the two systems using the same class
definition table. According to the GVW R2, about 7 percent of the weight data did not follow
linearity. This test is to investigate whether this 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
IRD
WPad
Vehicle Count
1000
800
600
400
200
0
c4
c5
c6
c7
c8
c9
c10
c11
c12
c13
1474 1293
12
167
21
1
33
172
53
3
1
2
WPad 1486 1283
18
150
21
1
34
171
51
3
1
2
IRD
c2
c3
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-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 weighpad 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 sideby-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 comparison 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 weighpad 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). Whatever 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 lowcost 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 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, National 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,” 6th International Conference on Managing Pavements, Brisbane, Australia,
2004.
[8] Y.H. Huang, Pavement Analysis and Design, 2nd 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,” Proceedings of the
Fifth IASTED International Conference on Circuits, Signals, and Systems (CSS 2007), pp. 233238, 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 87th 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-inMotion 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, 4th Ed.,
Brooks/Cole Publishing Company, Pacific Grove, CA, 1995.
[21] T. Kwon, “BullConverter: User Manual,” Transportation Data 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 weighpad data was successfully collected for a side-by-side comparison with the IRD system in this
site.
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.
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.
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
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.
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
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 Limits window.
B-3
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,35.7
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,35.9
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,36.1
14,1,11:25:30,5,67,16.6,4.3,36.1,4.0,,,,,,,,16.41,23.41,18.15,13.26,20.80,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,12,0,35.6
18,2,11:26:06,0,20,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,35.5
19,2,11:26:07,2,72,9.5,,,,,,,,,,,2.77,2.97,,,,,,,,,,,5.75,2,0,80,35.6
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,35.5
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,,,,,,,,,,,1.72,0.68,,,,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,12,0,45.4
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,50.7
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,,,,,,,,,,2.45,2.61,0.54,,,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,48.5
11,2,11:35:30,2,65,8.8,,,,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,42.9
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,41.6
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,41.3
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,44.2
29,1,11:40:39,2,64,9.6,,,,,,,,,,,2.45,2.18,,,,,,,,,,,4.63,2,0,80,44.3
30,1,11:41:07,0,66,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,9,0,46.2
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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,44.0
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
50,2,11:47:44,2,67,11.7,,,,,,,,,,,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,,,,,,,,,,,,,,,,,,,,,,,,0.00,15,8,0,50.3
C-4