U.S.

Department of
Transportation An Implementation Guide for Wayside Detector
Federal Railroad
Administration Systems

Office of Research,
Development
and Technology
Washington, DC 20590

May 2019
 NOTICE
This document is disseminated under the sponsorship of the
Department of Transportation in the interest of information
exchange. The United States Government assumes no liability for
its contents or use thereof. Any opinions, findings and conclusions,
or recommendations expressed in this material do not necessarily
reflect the views or policies of the United States Government, nor
does mention of trade names, commercial products, or organizations
imply endorsement by the United States Government. The United
States Government assumes no liability for the content or use of the
material contained in this document.

NOTICE
The United States Government does not endorse products or
manufacturers. Trade or manufacturers’ names appear herein solely
because they are considered essential to the objective of this report.
 REPORT DOCUMENTATION PAGE Form Approved
OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson
Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
May 2019 Guide - October 2017
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
An Implementation Guide for Wayside Detector Systems
6. AUTHOR(S) DTFR53-12-D-0004
Monique F. Stewart and Edward Flynn (FRA), Brian Marquis (Volpe), and Sharma & Task 29
Associates
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
REPORT NUMBER
Sharma & Associates, Inc. Volpe National Transportation System Center
5810 S Grant Street 55 Broadway
Hinsdale, IL 60521 Cambridge, MA 02142
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
AGENCY REPORT NUMBER
U.S. Department of Transportation
Federal Railroad Administration
Office of Railroad Policy and Development
Office of Research, Development and Technology
Washington, DC 20590
11. SUPPLEMENTARY NOTES
COR: Monique F. Stewart
12a. DISTRIBUTION/AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
This document is available to the public through the FRA website.
13. ABSTRACT (Maximum 200 words)
To promote wayside detector system implementation efforts, the Federal Railroad Administration (FRA) has constituted an FRA
Future Team - Technology Working Group to progress technology focused research activities to further its safety mission.
Under this initiative, FRA partnered with Metro North Railroad (MNR) and Long Island Rail Road (LIRR) to jointly identify
opportunities to enhance safety through the analysis of their existing wayside detection systems.
This guide includes brief descriptions of various wayside detection technologies, their capabilities and impact on rolling stock
performance and operating safety. The requirements for installation site selection, data communication and storage, system
thresholds and calibration/maintenance requirements.
The guide briefly reviews different wayside detector systems in North America’s rail network, including Acoustic Bearing
Detectors (ABD) (Trackside Acoustic Detection System [TADS®], Railway Bearing Acoustic Monitor [RailBAM®]), Automatic
Cracked Wheel Detector (ACWD), Dragging Equipment Detector (DED), Hot Box Detector (HBD)/Hot Wheel Detector (HWD),
Truck Bogie Optical Geometry Inspection (TBOGI), Truck Hunting Detector (THD), Truck Performance Detector (TPD), Weigh-
in-Motion Detector (WIM), Wheel Impact Load Detector (WILD), Wheel Profile Measurement System (WPMS), and Wheel
Temperature Detector (WTD) systems; while more focus targeted WILD as an example of the best practices in the implementation
of wayside detector systems.
14. SUBJECT TERMS 15. NUMBER OF PAGES
Bearing temperatures, calibration, Dynamic Ratio, DR, Hot Box Detectors, HBD, Truck Hunting 99
Detector, THD, wayside detector systems, wheel flats, Wheel Impact Load Detector, WILD, 16. PRICE CODE
Wheel Temperature Detector, WTD
17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT
OF REPORT OF THIS PAGE OF ABSTRACT
Unclassified Unclassified Unclassified
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18 298-102

i
 METRIC/ENGLISH CONVERSION FACTORS
ENGLISH TO METRIC METRIC TO ENGLISH
LENGTH (APPROXIMATE) LENGTH (APPROXIMATE)
1 inch (in) = 2.5 centimeters (cm) 1 millimeter (mm) = 0.04 inch (in)
1 foot (ft) = 30 centimeters (cm) 1 centimeter (cm) = 0.4 inch (in)
1 yard (yd) = 0.9 meter (m) 1 meter (m) = 3.3 feet (ft)
1 mile (mi) = 1.6 kilometers (km) 1 meter (m) = 1.1 yards (yd)
1 kilometer (km) = 0.6 mile (mi)

AREA (APPROXIMATE) AREA (APPROXIMATE)

1 square inch (sq in, in )2
= 6.5 square centimeters (cm ) 2
1 square centimeter (cm2) = 0.16 square inch (sq in, in2)
1 square foot (sq ft, ft )
2
= 0.09 square meter (m ) 2
1 square meter (m2) = 1.2 square yards (sq yd, yd2)
1 square yard (sq yd, yd2) = 0.8 square meter (m2) 1 square kilometer (km2) = 0.4 square mile (sq mi, mi2)
1 square mile (sq mi, mi ) 2
= 2.6 square kilometers (km ) 2
10,000 square meters (m2) = 1 hectare (ha) = 2.5 acres
1 acre = 0.4 hectare (he) = 4,000 square meters (m ) 2

MASS - WEIGHT (APPROXIMATE) MASS - WEIGHT (APPROXIMATE)

1 ounce (oz) = 28 grams (gm) 1 gram (gm) = 0.036 ounce (oz)
1 pound (lb) = 0.45 kilogram (kg) 1 kilogram (kg) = 2.2 pounds (lb)
1 short ton = 2,000 pounds = 0.9 tonne (t) 1 tonne (t) = 1,000 kilograms (kg)
(lb) = 1.1 short tons

VOLUME (APPROXIMATE) VOLUME (APPROXIMATE)

1 teaspoon (tsp) = 5 milliliters (ml) 1 milliliter (ml) = 0.03 fluid ounce (fl oz)
1 tablespoon (tbsp) = 15 milliliters (ml) 1 liter (l) = 2.1 pints (pt)
1 fluid ounce (fl oz) = 30 milliliters (ml) 1 liter (l) = 1.06 quarts (qt)
1 cup (c) = 0.24 liter (l) 1 liter (l) = 0.26 gallon (gal)
1 pint (pt) = 0.47 liter (l)
1 quart (qt) = 0.96 liter (l)
1 gallon (gal) = 3.8 liters (l)
1 cubic foot (cu ft, ft3) = 0.03 cubic meter (m3) 1 cubic meter (m3) = 36 cubic feet (cu ft, ft3)
1 cubic yard (cu yd, yd ) 3
= 0.76 cubic meter (m ) 3
1 cubic meter (m3) = 1.3 cubic yards (cu yd, yd3)

TEMPERATURE (EXACT) TEMPERATURE (EXACT)

[(x-32)(5/9)] °F = y °C [(9/5) y + 32] °C = x °F

QUICK INCH - CENTIMETER LENGTH CONVERSION

0 1 2 3 4 5
Inches
Centimeters 0 1 2 3 4 5 6 7 8 9 10 11 12 13

QUICK FAHRENHEIT - CELSIUS TEMPERATURE CONVERSION

°F -40° -22° -4° 14° 32° 50° 68° 86° 104° 122° 140° 158° 176° 194° 212°

°C -40° -30° -20° -10° 0° 10° 20° 30° 40° 50° 60° 70° 80° 90° 100°

For more exact and or other conversion factors, see NIST Miscellaneous Publication 286, Units of Weights and
Measures. Price $2.50 SD Catalog No. C13 10286 Updated 6/17/98

ii
Acknowledgements
This effort would not have been possible without the support of Metro-North Railroad and their
engineering staff. Their help and cooperation is greatly appreciated.

iii
Contents
Executive Summary ........................................................................................................................ 1
1. Introduction ...................................................................................................................... 2
2. Requirements for Implementation of a Wayside Detector System .................................. 4
2.1 Site Selection Criteria .................................................................................................. 4
2.2 Operating Parameters and Wayside Detector Technology.......................................... 5
2.3 Data Communication, Storage and Sharing ................................................................ 6
2.4 Test Requirements for Launching a New System ....................................................... 6
2.5 System Thresholds and Performance Monitoring ....................................................... 7
2.6 Developing an Action Plan for Triggered Events ....................................................... 7
2.7 Calibration and Maintenance Requirements ............................................................... 8
2.8 Training ....................................................................................................................... 8
3. Overview of Various Wayside Detector System Requirements ...................................... 9
3.1 Acoustic Bearing Detector (ABD) .............................................................................. 9
3.2 Automated Cracked Wheel Detector (ACWD) ......................................................... 15
3.3 Dragging Equipment Detector (DED) ....................................................................... 18
3.4 Hot Box Detector (HBD) & Hot Wheel Detector (HWD) ........................................ 20
3.5 Truck Bogie Optical Geometry Inspection (TBOGI)................................................ 23
3.6 Truck Hunting Detector (THD) ................................................................................. 26
3.7 Truck Performance Detector (TPD) .......................................................................... 29
3.8 Weigh-In Motion Detector (WIM) ............................................................................ 32
3.9 Wheel Impact Load Detector (WILD) ...................................................................... 34
3.10 Wheel Profile Measurement System (WPMS) .......................................................... 36
3.11 Wheel Temperature Detector (WTD)........................................................................ 39
4. Case Study: Wheel Impact Load Detector (WILD) ....................................................... 44
4.1 Functionality, Benefits and System Components...................................................... 44
4.2 Site Selection Criteria of WILD System ................................................................... 53
4.3 WILD System Thresholds, and Performance Monitoring......................................... 57
4.4 Calibration and Maintenance Requirements ............................................................. 64
4.5 Examples of Current WILD Experiences and Best Practices ................................... 65
Appendix A. WPD Calibration Record Requirements ................................................................. 79
Appendix B. MNR’s Faults Page with WILD Alert ..................................................................... 80

iv
Appendix C MNR’s Mechanical Maintenance Alert MMA # 292............................................... 81
Appendix D Cross-sections of HBD and HWD Systems ............................................................. 82
5. References ...................................................................................................................... 83
Acronyms and Abbreviations ....................................................................................................... 87

v
Illustrations
Figure 3-1. TADS® system, developed at the TTC [16] ............................................................... 10
Figure 3-2. An example of RailBAM® detector installed on the track [24] ................................. 13
Figure 3-3. Snapshot of the bearings fault signals generated by RailBAM® detector [24] .......... 13
Figure 3-4. A cup fault (left) and cone fault (right) identified by RailBAM® detectors .............. 14
Figure 3-5. Schematic plan of track structure (wide gauge, guard-rail) for ACWD site [30] ...... 15
Figure 3-6. An ACWD installed at the TTC by Tycho [30] (top), ACWD detector in the TTC is
spraying water couplant to conduct ultrasonic inspection on the wheels (bottom) [32]
.................................................................................................................................... 16
Figure 3-7. Schematic plan of ACWD site developed by Nordco [26] ........................................ 16
Figure 3-8. Another type of wheel crack detector deployed in Europe using Electromagnetic
Acoustic Transducers [9]............................................................................................ 17
Figure 3-9. An example of a Dragging Equipment Detector (DED) [3] ...................................... 19
Figure 3-10. A dragging component detected by DED system [10]............................................. 19
Figure 3-11. An example of HBD system (outer detectors) combined with HWD system (inner
detectors) [41]............................................................................................................. 21
Figure 3-12. An example of a hot journal flagged by an HBD detector with multi-scan system
(eight channels) [43] ................................................................................................... 21
Figure 3-13. TBOGI system to identify poorly performing trucks [8] ......................................... 24
Figure 3-14. Triple-box configuration of a TBOGI-HD system [47] ........................................... 24
Figure 3-15. Different types of geometrical errors of trucks which can be detected by TBOGI-
HD [46]....................................................................................................................... 25
Figure 3-16. THD system based on strain-gage detectors [6] ...................................................... 27
Figure 3-17. A sample history of THD alerts and info on a given rail car monitored through
InteRRIS®- EHMS system [29] .................................................................................. 28
Figure 3-18. Force-based TPD cribs and detectors laid-out along a given S-curve [53].............. 30
Figure 3-19. Flange and longitudinal creep forces acting against the rotational resistance between
truck and carbody [51] ............................................................................................... 31
Figure 3-20. WIM strain gages attached to the web of the rail on an instrumented track [9] ...... 33
Figure 3-21. Strain gages of WILD system attached to the inner sides of rail web [65] .............. 35
Figure 3-22. An example of WPMS system with laser triangulation and matrix camera [9]....... 37
Figure 3-23. Left: wheel profile measurements captured by WPMS; Right: brake pad
measurements by a WPMS detector [71] ................................................................... 37
Figure 3-24. A comparison between current wheel profile produced by WPMS (blue line) and its
respective theoretical profile (red line) [71] ............................................................... 38

vi
Figure 3-25. An example of WTD for brake effective tests (left); (right) Wheel temperature
variation on a train indicating braking condition [8] .................................................. 40
Figure 3-26. Wheel temperature on different cars of a test train in “Non-braking” status using
WTD [71] ................................................................................................................... 41
Figure 3-27. Wheel temperature on different cars of a test train in “Braking” status using WTD
[71] ............................................................................................................................. 42
Figure 4-1. Relationship between static loads and peak loads collected by a WILD system in a
Class I railroad network, classified for different types of rolling stock [75] ............. 46
Figure 4-2. Different components of a sample WILD system, based on strain gage technology [2]
.................................................................................................................................... 48
Figure 4-3. An example of WILD site components developed by L.B. Foster [59] .................... 48
Figure 4-4. Architecture of WheelCheX® as an example of an available WILD system ............. 49
Figure 4-5. Voestalpine’s WILD system using optic sensors underneath of the rails [10] .......... 50
Figure 4-6. An example of wheel tread irregularity detection concept of MERMEC’s WILD
system using accelerometers and strain gages [9] ...................................................... 50
Figure 4-7. Comparison of impact loads on wood and concrete ties at 40 mph, 100-ton cars, and
machined flats [69] (recreated chart).......................................................................... 54
Figure 4-8. Impact load scatter (36-inch diameter out-of-round wheel, single 140 mm divot, 100-
ton loaded car [69] (recreated chart) .......................................................................... 56
Figure 4-9. Impact load statistics (36-inch diameter out-of-round wheel, single 140 mm divot,
100-ton loaded car [69] (recreated chart) ................................................................... 56
Figure 4-10. Rate of WILD readings (per 1,000 wheels) and seasonal effect in North America
during 2003–2013 [8] ................................................................................................. 61
Figure 4-11. Impact of WILD installations on track cause related derailments [7] ..................... 63
Figure 4-12. Interdepartmental data flow for MNR's WILD system [1] ...................................... 68
Figure 4-13. Follow-up actions for a WILD alert at MNR [1] ..................................................... 69
Figure 4-14. Information/action sequence of WILD system at MNR [1]..................................... 70
Figure 4-15. An example of maximum loads time history of a 100-ton 4-axle car wheels—empty
and loaded conditions during 2004–2012 [7] ............................................................. 71
Figure 4-16. An example of AMS GUI details for a selected MNR train [1] .............................. 72
Figure 4-17. “AAR Condemnable” wheels for the US WILD systems (monthly count) [7] ....... 73
Figure 4-18. Seasonal trend on number of flagged wheels by MNR’s WILD system during the
fall season (particularly November) of 2010–2015 .................................................... 74
Figure 4-19. Comparison between number of flagged wheels by WILD and number of truing due
to “flat and shell spot” irregularities, collected for the given car series of MNR during
2010–2015 .................................................................................................................. 75

vii
Figure 4-20. Histogram of flagged wheels by WILD system for a given car of MNR commuter
RR during 2010-2015 ................................................................................................. 76
Figure 4-21. Mean load distribution of two selected cars (#XXXX on left and #YYYY on right)
based on WILD data collected during 2010–2015 ..................................................... 78

viii
Tables
Table 3-1. Summary of all wayside detector systems covered in this guide .................................. 9
Table 3-2. Summary of main requirements and specifications of TADS® (ABD) system........... 12
Table 3-3. Summary of main requirements and specifications of RailBAM® (ABD) system ..... 15
Table 3-4. Summary of main requirements and specifications of ACWD system ....................... 18
Table 3-5. Summary of main requirements and specifications of DE .......................................... 20
Table 3-6. Summary of main requirements and specifications of HBD and HWD systems ........ 23
Table 3-7. Summary of main requirements and specifications of TBOGI system ....................... 26
Table 3-8. Summary of main requirements and specifications of THD system ........................... 29
Table 3-9. Summary of main requirements and specifications of TPD system............................ 32
Table 3-10. Summary of main requirements and specifications of WIM system ........................ 34
Table 3-11. Summary of main requirements and specifications of WILD system ....................... 36
Table 3-12. Summary of main requirements and specifications of WPMS system ..................... 39
Table 3-13. Wheel temperature standard deviation associated with body rigging or hand brake
[51] ............................................................................................................................. 42
Table 3-14. Summary of main requirements and specifications of WTD system ........................ 43
Table 4-1. List of available WILD suppliers around the world .................................................... 49
Table 4-2. WILD system thresholds to maintain the reliability and precision level of the collected
data ............................................................................................................................. 51
Table 4-3. An example of measured and calculated wheel impacts at various speeds (source: CP)
[78] ............................................................................................................................. 55
Table 4-4. WILD indices for freight service [7] ........................................................................... 57
Table 4-5. MNR’s decision table for “Actionable” WILD data [1] ............................................. 58
Table 4-6. History of “Truing” and “Change-out” actions on the wheels of the same given car of
MNR commuter railroad during 2010–2015 .............................................................. 77

ix
Executive Summary
The Federal Railroad Administration’s (FRA) goal in supporting wayside detector system
implementation efforts is to promote improved safety. This goal can be achieved by monitoring
equipment performance through the appropriate and optimal application of automated inspection
technologies to monitor railroad rolling stock equipment performance and identify maintenance
needs proactively. The wayside detector technologies not only monitor rolling stock equipment
performance, the associated data from the wayside detector technologies can be used to predict
the future failures which may be prevented.
In 2015, FRA constituted the FRA Future Team - Technology Working Group (herein referred to
as the FRA Future Team) with the intent of progressing selected technology related research
activities that can further its safety mission.
Under this initiative, FRA partnered with Metro-North Railroad (MNR) and Long Island Rail
Road (LIRR) to jointly identify opportunities to enhance safety operations through the analysis
of existing wayside detection systems and related operational procedures. FRA’s Office of
Research, Development and Technology (RD&T) is also interested in extending such efforts
with other railroads in the future. Based on the lessons learned in the cooperative project with
MNR and general industry efforts in pursuing implementation of various wayside technologies,
the FRA Future Team has focused on developing this implementation guide which can help
railroads implement wayside detector systems.
This guide includes brief descriptions of various wayside detection technologies, which are
available in North America’s rail network to inspect and monitor rolling stock health status,
particularly in terms of wheel and truck related irregularities. It provides brief descriptions of
different recommendations for implementing a new wayside detector system including: site
selection criteria, data communication, test requirements, system thresholds, action plan for
triggered events, and calibration/maintenance requirements.
The guide briefly described and reviewed various wayside detector systems including Acoustic
Bearing Detector (ABD), Trackside Acoustic Detection System (TADS®), Railway Bearing
Acoustic Monitor (RailBAM®), Automated Cracked Wheel Detector (ACWD), Dragging
Equipment Detector (DED), Hot Box Detector (HBD), Hot Wheel Detector (HWD), Truck
Bogie Optical Geometry Inspection (TBOGI), Truck Hunting Detector (THD), Truck
Performance Detector (TPD), Weigh-In Motion Detector (WIM), Wheel Impact Load Detector
(WILD), Wheel Profile Measurement System (WPMS), and Wheel Temperature Detector
(WTD) systems. Each section includes a table of criteria which summarizes primary
specifications and requirements of respective systems.
As WILD is one of the most widely deployed systems, this guide contains a detailed discussion
on its functionality and benefits, site selection criteria, system thresholds and preference
monitoring, calibration and maintenance requirements, and some examples of WILD practices
experienced by selected North American railroads.

1
1. Introduction
Over the years, the Federal Railroad Administration (FRA) funded efforts to develop and
implement wayside detection systems for equipment performance monitoring, and, in some
cases, to promote and support deployment of detectors. These detectors are designed to reduce
risk in railroad operations by identifying poorly performing equipment before accidents occur.
In 2015, FRA constituted the FRA Future Team - Technology Working Group (herein referred to
as the FRA Future Team) to advance selected technology related research activities that can
further its safety mission.
Under this initiative, FRA partnered with Metro-North Railroad (MNR) and Long Island Rail
Road (LIRR), and jointly identified opportunities to enhance safety operations through the
analysis of existing wayside detection systems and related operational procedures. FRA’s Office
of Railroad, Development and Technology (RD&T) is also interested in extending such efforts to
other railroads in future.
Past efforts in using wayside detector devices and systems in support of mandated visual
inspections were stymied by the general lack of understanding and definition of the procedures
associated with the calibration and maintenance of these systems, as well as issues surrounding
the definition of maintenance limits, exception limits, data analysis procedures, etc.
To address these concerns, Sharma & Associates, Inc. (SA) was tasked by FRA to conduct a
review of the MNR's wayside systems, particularly focusing on Wheel Impact Load Detector
(WILD) implementation as integrated into operations and maintenance practices by MNR. SA
determined that MNR implemented well-thought out and integrated wayside detector systems
with the ability to enhance MNR's safety performance. The alarm criteria and resulting actions
have allowed MNR to effectively monitor and detect changes in equipment performance, and
take appropriate preventive measures to keep the rolling stock safely.
As a part of this project, SA was tasked with developing a Wayside System Implementation
Guide that could be used by other railroads that plan to install and implement wayside detection
systems. These systems are useful for fleet performance monitoring, improving operational
procedures, effective maintenance practices, and enhanced safety.
The contents of this guide are organized as follows:
• Section 2 provides a brief description of various major requirements for implementing a
wayside detector system along a given rail corridor. This section covers various criteria
and specifications including: site selection criteria, operating environment, data
communication and storage, test requirements for launching a new system, system
thresholds, action plan for triggered events, and calibration and maintenance
requirements.
• Section 3 briefly reviews wayside detector systems that are currently in use in North
America’s rail network to inspect and monitor rolling stock health status, particularly in
terms of wheel and truck related irregularities. This section covers various detector
systems including: Acoustic Bearing Detector (ABD), Trackside Acoustic Detection
System (TADS®), Railway Bearing Acoustic Monitor (RailBAM®), Automated Cracked
Wheel Detector (ACWD), Dragging Equipment Detector (DED), Hot Box Detector
(HBD), Hot Wheel Detector (HWD), Truck Bogie Optical Geometry Inspection

2
 (TBOGI), Truck Hunting Detector (THD), Truck Performance Detector (TPD), Weigh-In
Motion Detector (WIM), Wheel Impact Load Detector (WILD), Wheel Profile
Measurement System (WPMS) also known as Wheel Profile Detector (WPD), and Wheel
Temperature Detector (WTD) systems. Each section includes a table of criteria which
summarizes primary specifications and requirements of respective systems.
• Section 4 provides extensive details on the wheel impact load detector (WILD) system, a
technology widely deployed across the national rail network, including the users
experience and the best practices reported. WILD system characteristics detailed in this
section, include: system functionality and benefits, site selection criteria, system
thresholds and preference monitoring, calibration and maintenance requirements, and
some examples of WILD practices experienced by selected North American railroads.
• Appendices A through D provides some additional information regarding WPD, WILD,
HBD and HWD systems.
This guide will be updated as additional information on the wayside systems from the
participating railroads becomes available.

3
2. Requirements for Implementation of a Wayside Detector System
There are several varieties of wayside detectors in the rail industry that are being deployed by
railroads to monitor equipment performance and identify rolling stock defects and irregularities.
These detectors are designed to reduce risk in railroad operations by identifying poorly
performing equipment before accidents occur. The Association of American Railroads (AAR)
developed an Equipment Health Management System (EHMS) to house the data from a majority
of detectors. The AAR and the freight industry developed the criteria for their integration.
There are several criteria and features for installing a new wayside detector system that should be
considered to increase the likelihood of operating the detectors with higher productivity and
efficiency. This section reviews the primary requirements of wayside detector systems intended
for implementation along a given rail corridor. More details about functionality, benefits, and
particular requirements and characteristics of each wayside detector system are described in
Section 3.

2.1 Site Selection Criteria

The location of a new wayside detector system should be carefully evaluated and selected based
on the following criteria and respective system characteristics recommended by the vendor or
consultant firm [1, 2].

2.1.1 Track Characteristics

Track characteristics such as track quality and geometric specifications can be critical in
selecting the location of a new wayside detector system. Some of the detectors such as WILD,
TBOGI, and TPD systems may need specific track requirements and criteria to operate
effectively, such as the following:
• Track Components: Examples include only tracks with concrete crossties or ballasted
tracks, or welded rails, or a specific type of rail and fastening system.
• Track Geometry Requirements: Examples include no horizontal curvature, or
significant grades.
• Track Structure Requirements: Examples include that they may not be installed in a
tunnel, on a bridge, along a grade-crossing, or within the boundaries of switches, unless
specified and permitted by the manufacturer.
• Ride Quality: Some of the wayside detector systems may need to be installed along
locations to minimize the ambient noise level, vibration, etc., which may affect the
wayside system sensors.

2.1.2 Traffic and Speed Coverage

A wayside detector system should be located so that it has as much of the traffic flow past it as
possible to derive the maximum benefit from the installation. Some wayside detector systems
including WILD, WIM, HBD, etc. are able to capture traffic in both directions, while other
systems may be limited to only one direction due to operating practices, which may require dual
installations. The decision-making process for locating a system should also consider:

4
 • Operational speed of the rail vehicles passing along the given location
• Speed limit of the given track segment
• Minimum and maximum speed thresholds of the wayside detector

2.1.3 Location Accessibility and Maintenance Features

The prospective location of a wayside detector system should allow suitable access for
conducting inspection and maintenance activities. For detector systems that need immediate
action (follow-up inspection or repair action), there should be a rail facility located within a
reasonable distance from the proposed location of the wayside detector site so that a train crew
can quickly stop a train to inspect or set out the cars flagged by the wayside system. This is
highly dependent on the network layout and type of railroad service.
Other accessibility criteria, such as easy access to communication and power, security, right of
way aspects, and the history of neighboring interaction with railroad assets, should be also
considered. Also, for some of the wayside detectors such as HBD and WILD, it is typically
recommended to install these systems outside of train stop locations (e.g., away from yards,
sidings, stations, or hump yard boundaries) that tend to create high variability in operating
speeds. However, detectors such as WIM, or DED systems, which have less dependency on
speed, can be installed within the yard and station territories.

2.1.4 Other Criteria

Other criteria, such as avoiding interference with already existing equipment and facilities should
also be considered when selecting a location for a new wayside detector system. It may be
possible to integrate the proposed wayside detector with existing wayside systems (such as
WILD and HBD, or HBD and HWD systems), so this situation should also be considered.
Also, the location of a new wayside detector system should take into consideration any
limitations and requirements recommended by the vendor in terms of installation, maintenance,
and inspection to minimize rail traffic disruptions.

2.2 Operating Parameters and Wayside Detector Technology

The type of a new wayside detector system should be selected such that it can meet the expected
objectives of the new system on the rail corridor considering the operating parameters which are
practiced on that section. Several factors should be considered including, but not limited to, the
nature of the railroad (passenger, freight or commuter), operating rules, special instructions, train
frequency, carload volumes, and track specifications. For instance, the precision and resolution
level of WILD readings might be more important to consider for potential installation on a
commuter or a light rail system, in comparison to a heavy freight and intercity passenger corridor
with much higher axle loads.
The required time for replacing and repairing the components of a new wayside detector system
would be an important parameter for a congested rail corridor with tight train schedules, since a
long procedure for wayside detector maintenance may cause more delays and service
interruptions for congested corridors.
Some wayside detector systems may not be applicable for certain rail operating parameters. For
instance, the DED system is typically installed on the segments of a corridor with mixed

5
passenger and freight train operation practices [3]. WIM and THD, on the other hand, are
typically installed in corridors with more freight rail operations [4, 5, 6].

2.3 Data Communication, Storage and Sharing

The wayside detector system can collect a massive amount of data every day, based on the
volume of traffic and the type of detection system. This data should be regularly collected,
processed, analyzed, and reported to a decision center (train crews, dispatchers or other
authorized crews). Also, it should be clarified what type of data and when the processed data
should be archived and restored for any further investigation. Such procedures should be
designed well in advance and may need to be integrated with other local or interchangeable
database management systems, such as the InteRRIS® system, 1 EHMS, UMLER®, or other
similar systems.
There are varieties of communication means developed for different wayside detector systems,
such as [7, 8, 9, 10]:
• Global Positioning System (GPS)-based communication between the on-site equipment
and other remote equipment and office staff
• Optical fiber for faster and higher communication capacities
• Hard wired high-speed modems and/or wireless cellular connections for intranet and
internet based communication features
• Radio communication between system components and authorized railroad personnel
Also, note that automatic email messages, alerts, radio transmissions, and shop orders can be
created and sent to various departments, depending on the type of wayside detector and urgency
level of the alert. Reliable communication between different wayside detectors, other remote
equipment and staff is needed.
Based on the internal communication guidelines, data can be shared with information
technology, signaling, mechanical, operations, and track and maintenance groups. In case of
hosting other railroads traffic over a given wayside detector (freight, passenger, or commuter);
there should be an agreement between the parties on how and what detector data will be shared.

2.4 Test Requirements for Launching a New System

All components and features of a new wayside detector system should be tested before turning
the system on for normal operations. The pass-fail criteria or thresholds of each test (e.g.,
expected values) and the instructions for conducting the tests should be established and
communicated to the respective staff.
The components and features tested may include, but are not necessarily limited to:
• Trackside detectors and sensors such as rail sensors, cables and connectors, power
supply, train identification equipment

1
A central health monitoring system which shares nine different wayside detector systems including WILD between
all authorized railroaders such as Class I and rail car manufacturers.

6
 • Wayside signaling bungalows/huts, acting as a control room, containing computers,
power relays, battery backups, and/or communication equipment, configured with
analytical programs, connectivity tools and diagnostic algorithms
• System features and functionality such as train detection, train configuration (car,
wheelset and axle levels), train or vehicle speed, train weight, axle loads, and temperature
All results of component testing should be recorded and maintained for potential future reference
[1, 11]. If the system has a self-diagnosis feature, it does not dismiss the need for verifying the
given feature of the wayside detector system before engaging in daily operations. The system
should store the results of such a self-diagnosis.

2.4.1 Successful Tests

There is no further test required for a system component when it successfully passes its test,
unless it is recommended by the system vendor. However, the system maintenance procedure
should be followed according to the vendor requirements even if the initial test (or follow up test
if required by the vendor) is successfully passed.

2.4.2 Handling Non-Successful Test

Any fault discovered during the test procedure should be recorded and diagnosed using the Fault
Diagnostic Procedure (FDP), as recommended by the system vendor [1]. If the FDP does not
resolve the issue and the suspicious component or feature is not properly functional, then this
issue should be referred to the vendor to be properly addressed. The system should not be
operated without re-applying the failed test and obtaining a pass result after conducting the
appropriate action recommended by the vendor.

2.5 System Thresholds and Performance Monitoring

The data collected from the wayside detector system should be analyzed based on the thresholds
and decision criteria which have been recommended by the manufacturer. The system thresholds
should be defined in a way that it meets the FRA and AAR requirements.
Different performance indices are typically defined depending on the type of wayside detector,
including temperature based, load based, dimension based, laser or acoustic pre-determined
based indices.
The validity of thresholds selected for a detector system should be assessed to not only meet the
minimum safety requirements, but should also accommodate the capacity of the inspection and
maintenance facilities to repair the equipment flagged by the system. It is recommended to
closely monitor the performance of the wayside detector system for some period to evaluate the
impact of the system thresholds on the other parts of operations. More details about the specific
thresholds of each wayside detector system are provided in Section 3.

2.6 Developing an Action Plan for Triggered Events

Once a wayside detector system flags rail equipment (wheel, axle, journal, truck or car) with
measured values outside the acceptable threshold indices, there should be an appropriate action
to address the triggered incident recorded by the wayside detector. The respective actions can be
generated either automatically by the detector software or it can be developed by the respective

7
authorized parties (rail operator, car owner, etc.) as a result of further analysis. The analysis
should be reported in near real time for those irregularities and alarms which can endanger the
safety of rail vehicle operations. Some of the actions developed by wayside detector systems are
as follows:
• Stop operating the rail equipment immediately until further inspection is conducted
• Train speed reduction until further inspection is conducted
• Inspect the rail equipment as soon as practicable
• Check the flagged component during the next scheduled inspection or maintenance event
More details about follow-up actions for each wayside detector system are explained in the next
section.

2.7 Calibration and Maintenance Requirements

After launching a new wayside detector system, it should be calibrated and regularly maintained
based on Original Equipment Manufacturer (OEM) requirements and recommendations.
The calibration requirements and frequency of the calibration is typically based on the type and
functionality of the wayside detector system as well as the operating parameters. The calibration
procedure may need to be facilitated by additional programing, macros, or specific tools that the
manufacturer provided in advance. Also, some of the wayside detector systems are designed to
be automatically calibrated without using any extra procedure or a manually performed
calibration [4, 9, 10].
The calibration may need to be checked often to protect against conditions that can affect the
outcomes of wayside detector systems, such as environmental impacts, new types of cars passing
by the detector, track conditions, and aging of instrumented components. If the calibration
accuracy falls outside the accepted limits, then a re-calibration or maintenance action should be
conducted on the wayside detectors following the respective regulations of the rail operator [12].
Monitoring and analyzing different aspects of collected data by wayside detector systems can
also help respective authorities decide whether the calibration procedure needs to be adjusted to
improve the performance and productivity of wayside detectors or not. If the overall monitoring
and trending analysis of the detector system demonstrates an acceptable level of productivity and
effectiveness, then it can be concluded that the current calibration procedure was properly
designed. Otherwise, it is recommended to review the calibration procedure and modify the
sensitivity and decision trigger algorithm or replace certain parts of the sensors, if needed.

2.8 Training
To realize the full potential of the wayside detector system, the personnel who interact with it,
directly or indirectly, should be trained on how to install, function, inspect or maintain the
components of the system depending on the role and responsibilities. Generally, personnel from
signaling and communication, information technology (IT), mechanical, operations, and track
(infrastructure) should be engaged early in the planning and implementation stages. The training
framework and its contents should be developed based on the type of the wayside detector and
the railroad organization structure.

8
3. Overview of Various Wayside Detector System Requirements
This section reviews the most common wayside detector systems in North America’s rail
network that are used to inspect and monitor rolling stock health status, particularly in terms of
the wheel and truck related defects and irregularities. The review of each wayside detector
includes a table of criteria that presents primary specifications and requirements of the respective
system. Table 3-1 summarizes all wayside detector systems which were covered in this guide.
Table 3-1. Summary of all wayside detector systems covered in this guide
Wayside detector Functionality No. of available detectors
ABD (TADS®) Detecting the internal defects of wheel bearings long 19 (Freight: 19, Passenger: 0,
before they fail, using acoustic technology Commuter: 0)
ABD 20 (Freight: 20, Passenger: 0,
(RailBAM®) Commuter: 0)
ACWD Identify wheel flange cracks and internal defects by 2 (Freight: 2, Passenger: 0,
submerging the wheel tread through water Commuter: 0)
DED Detecting the components of rolling stock that are More than 1,000 DEDs (mostly
loose under moving trains along freight rail network)
HBD Detecting the wheel bearings defects, using an infrared More than 6,000 HBDs (mostly
thermal detection system along freight rail network)
HWD Detecting hot wheels due to locked or sticking brake NA (most of HWDs are integrated
shoes, using an infrared thermal detection system with HBD systems)
TBOGI Measuring the performance of car axles and wheel 28 (Freight: 28, Passenger: 0,
suspension using a laser-based technology combined Commuter: 0)
with a high-speed camera along a tangent section of
track
THD Evaluating truck hunting behavior (measuring hunting 94 (Freight: 90, Passenger: 4,
index) using strain-gages or laser-based technologies Commuter: 0)
TPD Evaluating the suspension performance of trucks along 14 (Freight: 11, Passenger: 0,
a S curve section of track (strain-gages or laser-based) Commuter: 3)
WIM Measuring the overload, side-to-side imbalance, or NA (mostly along freight rail
end-to-end imbalance condition of the cars network)
WILD Detecting the wheel defects (e.g., flat, shell, thermal 185 (Freight: 172, Passenger: 9,
cracks) by analyzing the wheel impact loads Commuter: 4)
WPMS Measuring wheel profile defects (e.g., flange height 15 (mostly along freight rail
and thickness, rim thickness), using laser and high- network)
speed camera)
WTD Detecting hot wheels (due to locked or sticking brake More than 700 WTDs (mostly
shoes) and cold wheels (inoperative brake system) along freight rail network)
using infrared scanning technology

3.1 Acoustic Bearing Detector (ABD)

The acoustic bearing detector system is based on an analysis of acoustic signature generated by
various bearing components of a bearing namely, cone, cup and rollers [13, 14]. A bearing fault
excites structural response of the bearing components that radiate sound with signature
characteristics of the bearing fault. By analyzing these sounds, the system can separate out
precisely which bearing component has the fault present. Further, the analysis also generates a

9
severity index for these faults allowing monitoring of the bearing if a maintenance or
replacement is warranted.
There are two types of ABD system available in the North America’s rail network, known as
TADS® and RailBAM®. More details on each specific system are provided in the following
sections.

3.1.1 Trackside Acoustic Detection System (TADS®)

As shown in Figure 3-1, the TADS® system consists of three sets of microphone arrays that span
approximately 25 feet along the track. The TADS® arrays are positioned on both sides of track,
at the approximate height of railcar bearings. The system should be installed on tangent track,
where the minimum speed of passing train is 40 mph. Also, the system should not be located
where trains may need to apply brakes, to reduce the interference associated with train noise that
may cause false positives [15].

Figure 3-1. TADS® system, developed at the TTC [16]

According to the Transportation Technology Center (TTC), the TADS® system can cover both
tapered roller bearings (AAR and metric sizes) and spherical roller bearings. The system is
capable of processing the following items on bearings [16]:
• High risk defects (unusual low frequency sound sometime called “growling”)
• Cup defects (spalling, brinelling, water etch)
• Cone defects (spalling, brinelling, water etch)
• Roller defects (spalling, brinelling, water etch, seams), and
• Loose cones

10
In most cases these potential defects and irregularities can be detected in a single pass between
20 and 60 mph. TADS® can also prioritize the defects it detects.
The flagged wheels with bearing defects can be relayed to the control center or other respective
authorities, using e-mail, page, fax, or modem connection. The system is Automatic Equipment
Identification (AEI) integrated and is also compatible with InteRRIS® system, according to the
TTC [16].
According to the AAR requirement, and to validate a new ABD detector, a teardown inspection
must be conducted on 50 bearings flagged by the new system, and the results should present at
least a 90 percent positive Level-1 defect rate as discussed below [15].
As outlined in AAR standard S-6000, a defect is Level-1 if any of the following conditions are
met [17]:
- Total spalled or water-etched area 1.5 square inches or more on any one cup or cone
running surface
- Total spalled or water-etched area 1 square inch or more on any one cup or cone
running surface and any spalled area on another running surface
- Any area of orange peel surface. Orange peel surface resembles the look and texture
of an orange
- Any loose component indication, such as:
o Cone back face wear > 0.010 inch
o Indication of turning on the journal for the bearing or its mate
o Oversize cone bore
o Mounted lateral > 0.030 inch
o Average cap screw torque < 50% of criterion in the Manual of Standards and
Recommended Practices, Section G-II
- AAR-condemnable peeling or smearing
- AAR-condemnable brinelling
- Lubrication failure as described in Section G-II, Recommended Practice RP-631,
“Bearing Failure Progression Mode LU”
- Any cracked or broken component
- Any AAR-condemnable roller defect
- Fails criteria for over-heated bearing teardown according to failure progression mode
analysis
- Any defect with a depth greater than 1/8 inch
- Fluting/arcing caused by electric current
As of March 2017, 19 TADS® detectors are in operation nationwide [18]. Note that more than
50,000 growler type severe bearing defects have been identified by ABD detectors in the U.S.
rail network and removed from service since 2007 [8].

11
Table 3-2 on the following page presents a summary of requirements and specifications of
TADS® system.
Table 3-2. Summary of main requirements and specifications of TADS® (ABD) system

3.1.2 Railway Bearing Acoustic Monitor (RailBAM®)

RailBAM® is other ABD systems developed by TrackIQ (now under Wabtec management) to
detect early and advanced bearing defects (including extended surface, cone, cup, roller and
audible faults) before they grow to a point that may cause service disruption or even derailment
due to journal failure [13, 19, 20]. As of March 2017, 20 RailBAM® detectors are in operation
nationwide and are integrated with the InteRRIS® [18].
RailBAM® uses acoustic measurements from a series of microphones to listen to passing trains
within the speed range of 15 to 80 mph [24, 25].

12
 Figure 3-2. An example of RailBAM® detector installed on the track [24]
RailBAM® uses two primary components, which are tie mounted auxiliary sensors, and a signal
processing electronics rack located in a wayside enclosure. The system can be integrated with an
automatic equipment identification (AEI)/radio-frequency identification (RFID) reader to
identify an exact car, axle, or wheel on a passing train. The generated alarms can be delivered via
SMS, email, or message to the train control centers [24].
The RailBAM® is operated based on identifiable sound characteristics produced by bearing and
wheel impact faults. A bearing fault stimulates a structural response of the bearing components
that emits sound characteristics of the bearing fault signature. Proprietary signal processing
techniques allow the bearing fault signal to be distinguished from the structural noise, enabling
fault identification and classification (Figure 3-3). It should be mentioned that RailBAM® can
also detect and locate wheel flats based on receiving the acoustic signature through the train
consist [23].

Figure 3-3. Snapshot of the bearings fault signals generated by RailBAM® detector [24]
The pro-active bearing monitoring of RailBAM® and other similar products can facilitate a more
systematic maintenance plan, providing the following benefits [25]:

13
 - Reduction of in-service failures
- Longer maintenance intervals, and
- Minimizing train service interruptions
The RailBAM® system can generate the trending history for bearing faults for further analysis
and decision making on the faulty car or bearings. Figure 3-4 presents examples of faulty
bearings (cup and cone faults) identified by the RailBAM® system.

Figure 3-4. A cup fault (left) and cone fault (right) identified by RailBAM® detectors
Note that the “De-cross” talking feature of RailBAM® system using the measured wheel-array
geometry and acoustic propagation physics can reduce, in software, the effect of a large fault on
one axle from the adjacent axles with small or nonexistent faults [27].

14
 Table 3-3. Summary of main requirements and specifications of RailBAM® (ABD) system

3.2 Automated Cracked Wheel Detector (ACWD)

ACWD is a wayside detector used to identify wheel flange cracks and internal defects by
submerging the wheel tread through water used as a couplant for ultrasonic detection over the
wheel rim [7]. The ACWD system typically requires a relatively stiff track structure (concrete
foundation or non-ballasted track) with certain track elements such as guardrails [30].
In addition to submerging the wheel tread through the water, ACWD may apply couplant
sprayed onto the lower surface of the rail wheel using a fine mist. The integrity of the wheel is
measured by ultrasonic transducers as the wheels pass through the inspection stations at speeds
from 5 to 15 mph [25, 26]. As shown in Figure 3-5 and Figure 3-6, the Tycho ACWD detector
system has been installed through a wide gage track structure because the ultrasonic equipment
should be able to inspect the wheel tread and rim as the wheels pass through. Thus, a guard rail is
provided on both inner sides of the track to protect the train against the derailment incident.

Figure 3-5. Schematic plan of track structure (wide gauge, guard-rail) for ACWD site [30]

15
 Figure 3-6. An ACWD installed at the TTC by Tycho [30] (top), ACWD detector in the
TTC is spraying water couplant to conduct ultrasonic inspection on the wheels (bottom)
[32]
Another type of ACWD detector, developed by Nordco for a Class I railroad, also uses ultrasonic
technology to detect wheel flaws and cracks through the wheel rim or tread.

Figure 3-7. Schematic plan of ACWD site developed by Nordco [26]

As shown in Figure 3-7, Nordco’s ACWD site consists of eight inspection stations on both sides
of the track (four stations per rail) with one ultrasonic test head per station. As the cars pass
through the inspection stations, the ACWD system sprays couplant onto the lower surface of the
wheel. Each ultrasonic station inspects one wheel in a four-wheel sequence, helping the system
to distribute the workload among different inspection stations and minimizing the risk of
inspection station failure [26].
After detecting a flaw or crack-related defect on a wheel, the system generates a real-time
assessment and may report the flaw types, sizes and locations that have been detected in and
across the tread and rim of the wheel [26].

16
Another wheel crack detection technology, developed by MERMEC in Europe, is a non-contact
detection system based on using Electromagnetic Acoustic Transducers installed along the track
without any requirement for wider track gauge or couplant (Figure 3-8). This ultrasound detector
sends acoustic Rayleigh waves through the wheels, which pass by the system at speeds up to 12
mph, and detects any cracks through the wheel tread, after processing the propagated surface
waves [9].

Figure 3-8. Another type of wheel crack detector deployed in Europe using
Electromagnetic Acoustic Transducers [9]
It should be mentioned that there are only a few ACWD systems in the U.S. rail network that are
in operation or under development by AAR, FRA, or Class I railroads, and none of them have
been integrated into the EHMS-InteRRIS® system yet [7].

17
 Table 3-4. Summary of main requirements and specifications of ACWD system

3.3 Dragging Equipment Detector (DED)

DED is a common wayside detector system that can protect rolling stock, cargo, and track
structures by detecting the components of rolling stock that are loose and dragging under moving
trains [7, 13]. There are several types of DED detectors designed by different manufacturers.
However, a DED system typically consists of physical sensors mounted on top of the track ties,
as shown in Figure 3-9 and Figure 3-10.
DEDs use acceleration sensors to detect dragging equipment. A typical DED system is capable
of detecting any strikes with accelerations up to 600 g [10]. Railroads can program the detectors
to send an alarm at whatever threshold deemed necessary [3]. The DED system can also be
equipped with a “Talker” feature to instantly announce the location of loose equipment on the
train, using radio communications with train crew [30].

18
 Figure 3-9. An example of a Dragging Equipment Detector (DED) [3]
Generally, when a DED alarm is generated by the system, another DED alarm is not to be
reported until the alarm contacts have closed and reopened and at least three axles have been
counted by the system since the previous alarm. Such a system configuration can prevent
multiple alarms being generated by one contact opening [35].

Figure 3-10. A dragging component detected by DED system [10]

The “Third Rail Fouler” detector is a particular DED system along electrified tracks with a third
rail power system. It can detect any components of the rolling stock penetrating the third rail
clearance envelope. The system is mainly applicable in urban transit (subway) and commuter rail
systems where freight trains may operate on the same track as commuter services. This detector
can identify any dragging object on freight trains that may cause damage to the third rail
equipment before it enters the commuter train territory [7].
More than 1,000 DED detectors are in-operation, nationwide, but they are not integrated through
EHMS-InteRRIS® system [7, 13].

19
 Table 3-5. Summary of main requirements and specifications of DE

3.4 Hot Box Detector (HBD) & Hot Wheel Detector (HWD)
Bearing failure is one of the major rolling stock problems for rail operators. It causes
approximately 20 percent of the $800 million wheel removals ($160 million), annually, in the
North American rail network [37]. Therefore, the HBD is a very common wayside detector
system to monitor and identify hot journal bearings using an infrared thermal detection system as
wheelsets pass over the detector [13].
A bearing can become overheated very quickly and may even burn off in just 1 to 3 minutes. On
the other hand, it may cool off at reduced speed or stop point before it can be confirmed by a
detector system [37]. Thus, the North American railroads have installed more than 6,000 HBD
detectors throughout their network to reduce the risk of bearing failures due to overheating [13].
As result, there is an HBD system every 25 miles, approximately, along Class I freight rail
networks [15]. Train accident rates caused by axle and bearing-related factors have dropped 81
percent since 1980 and 59 percent since 1990 due to the use of HBD detectors [38].
It is a common practice for railroads to combine the HBD with a hot wheel detector (HWD)
system to identify both hot journals (due to bearing defects) and hot wheels (due to unreleased
hand brake, automatic brakes not fully released and sticking brake shoes or disks) (Figure 3-11).
HBD and HWD use similar infrared scanning technology (and in some cases even
interchangeable detectors), though their decision algorithms and threshold criteria are different
for detecting bad bearings and brakes 2 [39, 40].

2
As of 2016, AAR has established S6031 for WTD detectors.

20
 Figure 3-11. An example of HBD system (outer detectors) combined with HWD system
(inner detectors) [41]

Figure 3-12. An example of a hot journal flagged by an HBD detector with multi-scan
system (eight channels) [43]
As shown in Figure 3-12, HBD system can be configured with multi-scan mode to improve the
reliability of identifying a hot bearing.
According to AAR’s interchange rules, each of the detector types have condemning limits that
permit a component such as a wheelset to be taken out of service and replaced when the
component meets or exceeds the noted condition. For a HBD [39]:
• The threshold is 76.7 °C (170 °F) above ambient and hotter than that side of the train, or
• 35 °C (95 °F) above the temperature of the mating bearing on the same axle
HBD systems utilize a series of rules and algorithms to verify that hot bearings are not the result
of heat generated by the sun, or by defective brakes generating additional heat. Bearings
temperatures are analyzed individually and compared with the train’s norm [15].
The HBD detection is evaluated based on three specific levels of bearing rules, in which level 1
represents the most severe condition of the bearings. Each railroad may define different rules and
criteria to evaluate its bearing conditions. For instance, Burlington Northern Santa Fe Railway

21
(BNSF) has a program for managing hot wheel bearings with more than 100 rules and more than
500 variations [15].
When hot journals are detected, the train should be inspected and will be operated at a reduced
speed, if necessary, until it reaches a set-out location where the car can be removed from the
train. When a hot wheel is detected, the flagged car is inspected, and then if the problem can be
remediated, the car will continue its operations. However, if a repair cannot be made on the car,
its brakes will be cut out and the car will be moved to the next available set-out location [15].
The maximum speed coverage of HBD and HWD systems varies from 100 to 300 mph based on
different manufacturer claims. Also, both systems can be equipped with the “Talker” feature to
directly inform the train crew of bearing and wheel problems, if integrated with an AEI reader
[10, 39].
There are certain conditions that should be considered when selecting a location for HBD and
HWD systems. Some of these primary requirements are [44]:
• Site should be on level and tangent track
• In an area where trains do not normally require heavy braking applications
• At least 300 feet away from any grade-crossings
• Away from track joints, switch and side tracks
• Passing trains should usually traverse with at least 10 mph along the selected location
• Track structure should be stable and well maintained. If needed, ties and fastening system
should be replaced, track should be tamped, and ties spacing should be reduced to
improve the track quality and stability level.
Depending on the manufacturer, HBD and HWD sensors should be carefully installed to measure
journal or wheel rim temperatures. For instance, according to a guide developed by Southern
Technologies Corporation (STC), HBD scanners should be aligned to scan the bottom 3.5 inches
of the bearing housing, about 7.25 inches from the gauge line. In addition, HWD scanners should
be aligned to scan approximately 4 inches above the rail, as demonstrated in Appendix D [44].
Note that the HBD and HWD systems are not integrated into the EHMS-InteRRIS® and
individual railroads and car owners utilize the data from these sites to monitor bearings and hot
wheels on their fleets using proprietary trending and threshold limiting analysis [7].
Recently, the AAR has developed standards for detecting cold and hot wheels and the HWD
systems will be covered under the same [41].

22
 Table 3-6. Summary of main requirements and specifications of HBD and HWD systems

3.5 Truck Bogie Optical Geometry Inspection (TBOGI)

TBOGI is a wayside detector system used to measure the performance of a rail car’s axle and
wheel suspension assembly, by using a laser-based monitoring system combined with a high
speed camera along a tangent section of track as a train passes over the detector (Figure 3-13)
[13].

23
 Figure 3-13. TBOGI system to identify poorly performing trucks [8]
The TBOGI can measure the lateral position (or tracking position) and angle of attack of each
axle of passing trains up to a maximum speed of 180 mph. In addition, the TBOGI-HD edition
can measure laterally unstable trucks due to hunting issues on passing cars, by converting the
maximum lateral distance to a measure of truck hunting, using a triple box configuration of
TBOGI, as shown in Figure 3-14 and Figure 3-15 [8, 42, 43].

Figure 3-14. Triple-box configuration of a TBOGI-HD system [47]

24
 Figure 3-15. Different types of geometrical errors of trucks which can be detected by
TBOGI-HD [46]
The TBOGI system alerts can be automatically set up based on different threshold
configurations. For instance, BNSF has implemented a three-level threshold rule for its eight
TBOGI sites which were installed through their network since 2006. These three level thresholds
are handled as follows [44, 45]:
• Level 3 (least severe): Automatically generated and automatic handling. Handled at
destination or during empty cycle.
• Level 2 (moderate level): Automatically generated and manual handling. Due to the
increased severity, manual handling is required to pick more specifically when and
where it would be best to inspect.
• Level 1 (most severe): Automatically generated and manual handling. This alert level
may require the train to be stopped immediately.
As of March 2017, 28 TBOGI detectors (12 of them are TBOGI-HD) are in operation
nationwide, and integrated within InteRRIS® system [13, 18].

25
 Table 3-7. Summary of main requirements and specifications of TBOGI system

3.6 Truck Hunting Detector (THD)

THD is a wayside detector to evaluate a truck’s ability to maintain the centerline of the rail at
relatively high speeds (greater than 40 mph), when operating empty or lightly loaded. There are
two types of THD available in the market. The first system is based on rail strain gages that
measure lateral wheel forces on tangent track, while the other system is based on laser
measurements of the angle of attack and the lateral position of an axle, similar to the TBOGI
system [7, 13, 15].
The force-based THD systems consist of a series of strain gages, which can be co-placed along a
WILD site. The THD site typically needs a 30 to 50-foot array of strain gages welded to the web
of the rail, which measures the lateral forces created by running car wheels on the rail (Figure
3-16). The force data gathered from each wheel are evaluated using an index value (Hunting
Index), which is used to identify defects depending on the acceptable thresholds and criteria
configured for the system [6, 15].

26
 Figure 3-16. THD system based on strain-gage detectors [6]
On the other hand, the laser based THD system consists of three modules. Each module contains
an emitter and receiver for sending and receiving a laser beam. The modules are approximately
10 feet apart and located just above the head of the rail. The collected data of a laser based THD
is generated based on the speed, amplitude, and frequency of the laser beam as it is received [15].
The THD condemnable thresholds are typically evaluated based on a truck Hunting Index (HI)
that should be less than 0.20 [29, 50]. The HI uses an algorithm based on correlated lateral and
vertical data received from the THD detectors and is typically scaled between 0 and 1; though
under certain conditions the HI can be negative, too. An HI less than 0.1 demonstrates a stable
truck, while any values above 0.65 indicate a truck with very poor hunting stability [51, 52]. HI
with values higher than 0.1 show a degradation trend through the truck suspension system [51].
As an example, Figure 3-17 demonstrates history of THD alerts on a given car.
When a car is flagged by THD, different levels of inspection or maintenance actions may be
required depending on the HI thresholds defined, including adherence to AAR Interchange Rule
46, Section H. For instance, if the HI is not beyond the condemning thresholds, the flagged car
may continue its operations up to the destination or unloading location; otherwise it should be
only operated up to the next set-out point [52].

27
Figure 3-17. A sample history of THD alerts and info on a given rail car monitored through
InteRRIS®- EHMS system [29]
The maintenance and calibration requirements of THDs, are as follows [21]:
- “THDs must be maintained such that each rail has at least 70% of both vertical and
lateral circuits active (so as to determine L/V ratios at that rail at a crib). If less than
70% of both vertical and lateral cribs are active on either rail, then the data from
that site does not meet the validation requirements.
- The average vertical weight for all wheels measured must be calculated for each
active circuit. The range (maximum-minimum) of these average weights for a rail
must be less than 15 kips for any train set containing 50 or more axles. If the range is
greater than 15 kips, then data from that rail does not meet the validation
requirements.
- The average lateral load for all wheels measured must be calculated for each active
circuit. The range (maximum-minimum) of these average weights for a rail must be
less than 6 kips for any train set containing 50 or more axles. If the range is greater
than 6 kips, then data from that site does not meet the validation requirements.”
AAR standard S-6101 also lists the following data requirements for THD calibration [21]:

28
 - “The range of average weight variation for each rail for each train must be provided
with the train data set.
- The range of lateral load variation for each rail for each train must be provided with
the train data set.
- The percentage of active L/V circuits per rail must be provided with the train data
set.”
The typical installation of strain-gage THD systems requires concrete ties along the measurement
zone and the approaching segments. Also, changing tie spacing may be required to increase the
measurement accuracy level. The THD system may require sensors to detect and count the axles
or read the AEI/RFID tags. Typically, a central location server (in a bungalow) is located next to
the THD sensors to store and compare the collected data to the defined thresholds and generate
alarms and send them to the authorized users [47].
As of March 2017, 94 THD detectors are in-operation nationwide, which were integrated under
the InteRRIS®/EHMS system [18]. Note that since implementing THD systems in the U.S. rail
network, more than 20,000 freight cars were flagged by THDs, and these cars were repaired [8].
Table 3-8. Summary of main requirements and specifications of THD system

3.7 Truck Performance Detector (TPD)

TPD is a wayside detector system, which evaluates the suspension performance of trucks by
measuring the vertical and lateral forces generated by the wheels as a car moves over the
detectors that are placed along instrumented track [7, 13].

29
Similar to THD, TPD is also categorized on two different technologies: a strain-gage (force-
based) system and a laser-based system [15]. In force-based TPDs an array of strain gages are
installed throughout an “S” curve segment of track with 4–6 degrees of curvature, including six
to eight cribs of strain gages, in which two (or three) of them are placed in the constant radius
portion of both curves and two in the tangent section [15, 52], as shown in Figure 3-18. Each crib
measures both lateral and vertical forces applied to the rail, which are converted to an index
value to measure the risk severity of the defects related to the truck performance along the curves
[15]. The angle of attack is also measured using the time differential between the high and low
rails, with different levels expected in the radial segments of track versus the tangent segment
[47].
The laser-based TPDs consist of one module, an emitter and receiver, which generate and receive
a laser beam, similar to the TBOGI system and the laser-based edition of THD. The elapsed time
between sending and receiving the laser beam is compared and evaluated based on a known
distance from module to reference rail. The difference between the measured value and the
benchmark value is used to evaluate the truck performance. Unlike the force-based TPD system,
the laser-based THDs should be installed along a tangent segment of the track [15, 43].

Figure 3-18. Force-based TPD cribs and detectors laid-out along a given S-curve [53]

The TPD wayside detector can flag cars with the following defects [54, 55]:
• Worn friction wedges
• Broken suspension springs
• Twisted car bodies
• Mismatched side frames
• Hollow/Worn wheels
• Tight side bearings
In the force-based type of TPD, the system detects the truck issues based on calculating the
gauge widening forces of crib pairs for each truck along the circular curve and the mean force for
the entire train. The condemnable levels of TPD are classified based on L/V (TPDL) and gauge
spread (TPDG) thresholds, in which the first one deals with lateral over vertical force

30
measurements; while the latter refers to truck performance gauge (in other words the
measurement of the horizontal force of the truck against the inside rails, pushing them outward)
[50]. According to AAR, the condemnable levels of TPD systems are defined as below [18]:
• TPDL: LAHRLV ⃰ 3≥ 1.05 twice in 12 months
• TPDG: Gauge Spread ≥ 28+ kips (by degree curve) twice in 12 months
To assess if the TPD system produces reliable data, the requirements are that the absolute value
of average truck gauge spread force for a train (TPDG), which is calculated at each two cribs in
the curves, or along the tangent segment, be greater than or equal to 10 kips. Further, the lower
mean truck gauge spread force calculated at one crib for the train should not be less than 50
percent of that of the higher mean truck gauge spread force measured at the other crib for that
train. In addition, the lateral force of each single wheel at any crib, TPDL, should not be greater
than 45 kips. Also, the standard deviation of lateral force for a train at any crib should be greater
than zero but less than or equal to 8 kips. Similarly, the standard deviation of vertical force for a
train at any crib should be greater than zero [21].
According to research conducted by the TTC, most poor performance trucks of freight cars
detected by TPD are caused by high rotational resistance at the truck-body interface, as shown in
Figure 3-19 [51].

Figure 3-19. Flange and longitudinal creep forces acting against the rotational resistance
between truck and carbody [51]
It should be noted that the truck-car body rotational resistance can be caused by [51]:
• High friction coefficients on:
o The horizontal surface of the center bowl
o The side wall of the center bowl
• Body twist (causing high side bearer loads and off-center vertical load distributions)

3
LAHRLV: Lead Axle High Rail Lateral over Vertical ratio (L/V)

31
 • Twisted center plates across a wagon body
• Non-flat (concave) center bowls and center plates (as a result of carbody twist)
As of March 2017, 14 TPD detectors are in operation nationwide. Also, the TPD system has
been integrated through the InteRRIS® database management system [18]. It should be
mentioned that since implementing TPD systems in the U.S. rail network, approximately 5,000
freight cars have been flagged by TPDs and repaired [8].
Table 3-9. Summary of main requirements and specifications of TPD system

3.8 Weigh-In Motion Detector (WIM)

WIM is a wayside detector system that can measure the weight of a rail vehicle (vertical load)
and analyze any overload, side-to-side imbalance, or end-to-end imbalance condition of the cars
while they are moving in the 1–180 mph speed range, depending on the type of WIM system.
WIM can be combined with a WILD system or it can be installed as a stand-alone detector [5, 9].

32
Figure 3-20. WIM strain gages attached to the web of the rail on an instrumented track [9]
WIM should be integrated with an AEI reader to identify the car with imbalance or over-loaded
axles and automatically report the problem. There are certain weight related alarms that can be
reported depending on the WIM system characteristics, but some of the common reportable
problems are as follows [4]:
• Axle overload
• Transverse axle imbalance
• Transverse load imbalance
• Longitudinal load imbalance
The WIM systems can be developed with self-diagnostic capabilities in a way that if any major
part of the system fails (sensors, measurement equipment), the system can detect such problems
and send a warning alarm to the system users [5, 56].
According to Union Pacific Railroad (UP), during July 2005 and September 2006, more than 11
million vehicles were monitored by 13 WIM systems on UP main tracks. Out of 11 million
records captured by WIM systems, only 439 vehicles exceeded the max loading capacity by
more than 10 percent [57].
Table 3-10 presents a summary of main requirements and specifications of a WIM system.

33
 Table 3-10. Summary of main requirements and specifications of WIM system

3.9 Wheel Impact Load Detector (WILD)

The WILD system is one of the most common wayside detectors. It is used to identify the wheels
with potential tread defects such as flat spots, out-of-rounds, built-up treads, and shells that result
in high impact loads, which cause damage to the vehicle and truck components, as well as to the
track structure. Different WILD systems were manufactured under different commercial names
such as WILD [4, 9, 59, 60], WheelCheX® [12], MultiRail® WheelScan [64], WCM® [65], and
ATLAS FO [66].

34
 Figure 3-21. Strain gages of WILD system attached to the inner sides of rail web [65]
As of March 2017, more than 185 WILD detectors are in operation nationwide and have been
integrated through the InteRRIS® system [18]. Table 3-11 presents a summary of the main
requirements and specifications of the WILD system. More details about WILD systems are
provided in the Section 4.

35
 Table 3-11. Summary of main requirements and specifications of WILD system

3.10 Wheel Profile Measurement System (WPMS)

WPMS, also known as Wheel Profile Detector (WPD), is a wayside detector system used to
determine wheel wear, flange height and thickness, rim thickness, and hollow tread problems by
using a laser-based scanning system and capturing profile pictures with high-speed digital

36
cameras (Figure 3-22) [7, 9, 13]. The system has not been integrated into the EHMS system yet
[7], and as of March 2017, 15 WPMS detectors are in operation nationwide [18].

Figure 3-22. An example of WPMS system with laser triangulation and matrix camera [9]
There are various types of high-speed camera and laser-based measurement tools which were
developed by different manufacturers using digital image processing techniques to measure the
wheel profile and brake shoe pad or disk thickness [52].
The Fully Automated Car Train Inspection System (FactIS) is a machine vision inspection
technology that uses high-speed cameras and strobe lights installed next to the tracks to capture
images of wheels and brake shoes. These images are stored and analyzed in terms of the wheel
flange and rim irregularities, as well as evaluating the top and bottom of brake shoe thickness, to
determine the extent of uneven shoe wear [52, 67]. Figure 3-23 presents two sample images of
wheel profile and brake pad measurements captured by WPMS.

Figure 3-23. Left: wheel profile measurements captured by WPMS; Right: brake pad
measurements by a WPMS detector [71]
Once the images of wheel profile are produced by WPMS, the details of wheel components in
the images are analyzed and compared with the theoretical profile of the reference wheel, and
any irregularities are identified by the system (Figure 3-24).

37
 Figure 3-24. A comparison between current wheel profile produced by WPMS (blue line)
and its respective theoretical profile (red line) [71]
Depending on the severity level of irregularities measured for wheel profile criteria, the flagged
wheel can be immediately inspected and fixed, before reaching the maintenance or condemning
limits [47]. It should be mentioned that the WPMS systems are classified based on low speed
operations (0–7 mph) and high-speed operations (0–75 mph) technologies [9], though it is more
practical to capture the wheel profiles at normal train speeds [69, 70].

38
 Table 3-12. Summary of main requirements and specifications of WPMS system

3.11 Wheel Temperature Detector (WTD)

WTD is a wayside detector system used to scan the outer surface of railcar wheels and record
temperature for each wheel passing the detector by using infra-red scanning technology. By
determining the temperature of wheels, it can indicate whether brakes applied when they should
not or applied when they should. As of 2013, approximately 700 WTD are installed in the freight
network [7].
The WTD system uses similar technology and algorithms as the HBD system to analyze the
wheel temperature distribution throughout the train. As result, stuck brake shoes, unreleased
hand brakes, and inoperative valves can be identified using WTD. These are found from wheel
temperatures that are greater than (or less than) the wheel temperatures in the rest of the train
(Figure 3-25) [8, 54].

39
Figure 3-25. An example of WTD for brake effective tests (left); (right) Wheel temperature
variation on a train indicating braking condition [8]
According to AAR standard S-6031, the WTD should be able to capture the cold and hot wheels
on a car based on evaluating four different criteria as below [41]:
• Truck temperature ratio: which can be determined based on Equation (1)
• Wheel temperature: the absolute or average temperature value of a given wheel
(depending on the condition that is assessed)
• Sufficient wheel heat for testing: means the wheel heat is high enough to ensure that the
train is braking and thus, a lower temperature condition of particular wheels demonstrates
a poor braking performance.
• Sufficient lack of wheel heat for testing: means the wheel heat is low enough to ensure
that the train is not braking and thus, a higher temperature condition of particular wheels
demonstrates a poor braking performance.
The truck temperature ratio is a primary parameter which can be used in combination with other
criteria to detect hot or cold wheels through a train. This parameter can be determined using
Equation (1) [41]:

Where
𝑇𝑇𝑡𝑡𝑡𝑡 = Truck temperature ratio
𝑇𝑇𝑤𝑤 = Wheel temperature from wayside detector
𝑇𝑇𝑒𝑒 = Expected wheel temperature
𝑁𝑁 = Number of wheels on a given truck
For instance, according to AAR standard S-6031, an inoperative brake from insufficient brake
application can be determined on a car, under one of the following circumstances [41]:

40
 • Any truck on the car has a truck temperature ratio less than 30 percent and sufficient
wheel heat for testing exists as described in the AAR standard S-6031.
• One or more wheel temperatures on the car are less than or equal to 70 °F and more than
three standard deviations below the average for the train side, and sufficient wheel heat
for testing exists as described in the AAR standard S-6031.
Another example of WTD practices would be to determine an inoperative brake from insufficient
brake release on a car under one of the following circumstances [41]:
• Any truck on the car has a truck temperature ratio greater than or equal to 200 percent,
and sufficient lack of wheel heat for testing exists as described in the AAR standard S-
6031.
• The average value of the wheel temperature minus the expected wheel temperature,
divided by the inter quartile range of the train side, for all wheels on any single truck on
the car is 3.0 or more.
• One or more wheel temperatures on the car are greater than or equal to 200 °F and more
than three standard deviations above the average for the train side.
Figure 3-26 and Figure 3-27 present wheel temperatures of two test trains in “Non-braking” and
“Braking” status, which were measured by a WTD system on Class I main tracks through
research sponsored by FRA [71]. As shown in Figure 3-26, there are a few cars with high
temperatures over the average temperature of the given train that was supposed to be in non-
braking status, demonstrating poor braking performance on these detected cars. On the other
hand, Figure 3-27 shows certain cars with lower wheel temperature than the average temperature
of the given train that was supposed to be in braking status, demonstrating inoperative brake
components or a brake system malfunction on these cars [52, 71].

Figure 3-26. Wheel temperature on different cars of a test train in “Non-braking” status
using WTD [71]

41
 Figure 3-27. Wheel temperature on different cars of a test train in “Braking” status using
WTD [71]
According to another study conducted by the TTC, the average wheel temperature is associated
with the air brake valve performance and its major components as presented in Table 3-13.
Table 3-13. Wheel temperature standard deviation associated with body rigging or hand
brake [51]

As presented in Table 3-13, the largest standard deviation, 88 °F, is associated with the valve of
the brake system, body brake rigging or hand brakes [51].
To detect cold wheels (and trucks), the WTD systems should be installed on or at the end of a
descending grade where the brakes should be applied to a sufficient degree and long enough
period to generate heat between wheels and brake shoes or pads.
To detect hot wheels due to stuck brakes or hand brakes left applied, the system should be
installed far from any downgrades or any track segments requiring a normal brake application
[41].

42
Table 3-14. Summary of main requirements and specifications of WTD system

43
4. Case Study: Wheel Impact Load Detector (WILD)
WILD is used in this guide as one of the best documented examples of wayside detector systems
that can improve the safety of rail operations, as well as enhance the maintenance practices of
both rolling stock and track structure components. This section reviews more details and
information regarding the WILD system components, implementation criteria, system thresholds,
and WILD calibration and maintenance requirements.

4.1 Functionality, Benefits and System Components

The WILD system can provide an analysis of several types of loads that can be considered in
track structure and wheel-rail interaction analysis [75]:
• Static: The weight of the rail vehicle when it does not move.
• Quasi-static: The combined static load and the effect of static load at speed, independent
of time parameter. The quasi-static load can be measured in curved tracks when the train
generates additional loads onto the rails due to centripetal force and frictional or creep
forces.
• Dynamic: The additional load due to effects of wheel/rail interaction parameters
including time-dependent track component response as well as inertia, damping, stiffness,
and mass variables.
• Impact loads: Additional loads due to wheel irregularities (such as shelling, flat spots and
built up tread) that also create high loads on the track structure.

4.1.1 WILD System Functionality

The main purpose of the WILD system (and other similar systems) is to detect and monitor the
dynamic and impact loads of the rail vehicles while they are moving on the track, though they
can collect the static and quasi-static loads as well.
When a wheel irregularity generates a vertical load that is high compared to the normal wheel
loads (e.g., higher than 90 kips), it is considered to cause damage to equipment and track
structure components. In extremely high load cases (≥140 kips) it may not be safe to continue
operations at normal speed and the train might be directed to proceed at a reduced speed to the
nearest siding or yard to set out the car with high impact wheels.
The high impact loads (peak loads) are typically generated by wheels that have a flat spot on the
tread surface, which are often caused by a hand brake not released, or by locking brakes (slip-
slide). Since the wheel cannot rotate while the hand-brake is engaged (or locking brakes), the
wheel-rail friction can deform the surface of the wheel tread to become flat instead of round [73].
In addition to a flat spot, other wheel irregularities such as shell spots, and high flange may also
cause high impact loads.
The first application of the WILD system in the U.S. rail environment was in the early1980s on
the Northeast Corridor (NEC) when Amtrak inspectors were confronted with numerous
transverse rail seat cracks on their relatively new concrete crossties. The investigation revealed
that these cracks were caused by high peak vertical loads (up to100 kips) generated by wheel
tread irregularities (long wavelength out-of-round conditions) on Amtrak cars in such a way that

44
these irregularities could not be detected through normal visual inspection or regular wheel
geometry inspections. As a result, Amtrak and FRA funded a project a to develop the first WILD
system to be installed on NEC near Edgewood, MD, to detect wheels with high loads to prevent
further potential damage to the track structure, including tie life cycle and wheelsets [77].
After the WILD system implementation on the NEC corridor, Amtrak reported a significant
reduction in axle bearing issues (from once a month to once every 6 months), which might be
caused by high impact loads. Consequently, accidents due to bearing failures were decreased
along this corridor as well [78]. This successful WILD implementation created an impetus in the
industry to adopt the detector system for wide freight network use led through further research
by the AAR in 1990s.
High WILD loads can be presented as forces (kips or kN) or as Impact Factor or Dynamic Ratio
(DR) (peak load divided by static load). The peak loads are typically within a range of 50 up to
140 kips, while the impact factor is typically calculated between 1 and 5. For instance, an impact
factor of 2 means the peak load of the given rail vehicle is twice the static wheel load, e.g., 60-
kips peak load for a 30-kips static load.
Figure 4-1 demonstrates an example of peak loads versus the nominal loads collected by a WILD
system installed in a Class 1 railroad network. The various impact factors are shown as lines
demarcating the peak to static load ratios for the data. The percentages indicate the cumulative
number of ratios that appear in that zone and the lower zones. However, the 0.4 percent number
shown in the figure is different because it is the number of ratios appearing at or above an impact
factor of 5 in this zone, rather than the cumulative number of ratios shown elsewhere in the chart.
Note that since 2002 all WILD detectors in the U.S. rail network are included through a central
data repository system known as Equipment Health Management System (EHMS) that allows
sharing the collected data among the authorized freight railroads as well as rail shippers, rail
vehicle manufacturers, and car owners for further maintenance decision making [1].

45
Figure 4-1. Relationship between static loads and peak loads collected by a WILD system in
a Class I railroad network, classified for different types of rolling stock [75]

4.1.2 WILD System Benefits

There are several benefits provided by a WILD system including improvements to the following
[59, 65]:
• Wheel tread life
• Ride quality
• Fleet availability
And, reduction in the following:
• Rail fatigue
• Concrete and wood tie damage
• Bearing damage
• Truck damage
• Noise level
• Service interruptions
Although there are always only a few wheels with high impact loads in service, even a small
percentage of high impact wheel loads can cause a big influence on the operating costs and
safety aspects. For instance, a study revealed that high impact loads may increase the surface
crack growth rate on a rail by approximately 100 times compared to low impact loads [76]. Also,

46
according to another study, there are approximately 2,400 wheelsets per day in North America in
which there is a significant impact difference between the 2 wheels on the same wheelset [73].
Such a noticeable difference between two wheel loads may indicate that there is a potential
wheel defect on the wheel with the higher impact load.
The WILD system can not only capture high impact loads, it can also improve system efficiency.
For instance, it is indicated by MNR, a commuter rail service of Greater New York, that
implementation of four WILD systems effectively resulted in a 28 percent reduction in overall
impact loads, with the average DR dropping from 2.5 to 1.8 [68]. Another study conducted in
Australia revealed that installing a new WILD system could provide a reduction of 90 percent in
serious wheel related-irregularities during the first 6 months of WILD operations. The same
study also demonstrated that WILD detectors may provide a far earlier warning of bearing failure
than thermal ‘hot-box’ systems [77].
There are various wheel tread irregularities that can be detected by WILD systems, such as flat
and shell spots. In addition, vertical split rims (VSR) and shattered rims which can cause high
wheel loads and even derailment, are among those wheel irregularities that can also be detected
by WILD systems [73, 78]. For instance, in a study conducted by TTC, 24 broken wheels were
examined and it was confirmed that 71 percent of the time, the wheel failure was caused by a
VSR defect. Also, 12 out of these 24 broken wheels had historical WILD data in which 50
percent of them (6 out of 12 wheels) had impact loads that exceeded 90 kips prior to failure [78].
From a financial standpoint, WILD systems can have a significant impact on maintenance and
repair costs, due to the fact that wheel-related issues are a major cause of train accidents [78, 82].
According to MNR, applying one WILD system in its network increased the service life of its
wheelsets from 2.5 to 3 years. Such improvement could provide an annual savings of $1.6
million, or $16 million over 10 years, just in terms of wheelset maintenance cost (excluding any
benefits from reduction in damage costs to track and vehicle trucks) [68].
In Australia, the National Rail (NR) operator (currently known as Pacific National) reported that
their single WILD system could save between $1.2 to $2.8 million each year (1999-dollar value)
just in reduction of wheelset failure cost [77].

4.1.3 Major Components of a WILD System

Generally, a WILD system consists of three major components [78]:
• Instrumented tracks
o Train presence sensors (such as accelerometers, strain gages, optical sensors or a
combination of them), RFID or AEI tag reader
• Data logging and processing equipment
o Single embedded controller and signal conditioning unit running impact load
analyzer software
• Communication equipment
o Wired or wireless data transfer equipment (modem, fiber optic cables, GPRS,
etc.), power back up (solar battery, DC or AC power) and calibration
setup/features (may or may not be included)

47
As shown in Figure 4-2, a WILD system gathers signals from sensors (e.g., strain gages installed
on the web of the rail in a series of consecutive tie cribs) triggered by passing wheels, sampling
approximately 7 to 10 percent of each wheel circumference per crib [77]. The number of cribs
instrumented control the so-called “coverage” of various wheel sizes which determines the
probability of detecting a wheel irregularity as the wheel passes a WILD site.
All WILD sites are typically equipped with a technical (control) room next to the measurement
zone (instrumented track and AEI tag reader) to collect and analyze the respective data, and then
to transmit them to the central office or train crew (Figure 4-3).

Figure 4-2. Different components of a sample WILD system, based on strain gage
technology [2]

Figure 4-3. An example of WILD site components developed by L.B. Foster [59]
There are various types of WILD systems developed by different manufacturers or suppliers
around the world using different technologies and components. Table 4-1 presents the main
suppliers of different WILD systems.

48
 Table 4-1. List of available WILD suppliers around the world
Supplier Product Headquarter Notes References
Ansaldo STS WILD Italy Mainly provided in Europe [4]
DeltaRail 1 WheelCheX® UK Mainly provided in UK, Spain, and [67, 68]
US
L.B. Foster® WILD USA Primary manufacturer of WILD [59, 74]
Salient Systems systems in North America (over 190
systems installed in the US)
MERMEC WILD Italy Mainly provided in Europe [9]
NagoryFoster WILD India - Mainly provided in India and [80]
South-Asia
- Developed based on Salient’s
Technology
Progressive Rail WILD USA Mainly provided in US (over 15 [63]
Technologies 2 systems installed)
Schenck Process MULTIRAIL® Germany Mainly provided in Europe [61, 64]
WheelScan
Track IQ WCM® Australia Over 50 systems installed, worldwide [62, 77]
Voestalpine ATLAS FO Germany Mainly provided in Europe [10, 66]
1: Currently, managed by Vortok, part of Pandrol
2: Currently, part of “International Engineering and Global Connections”

Figure 4-4 presents the architecture of one of the WILD systems (WheelCheX®) including the
trackside equipment as well as communication and control center tools.

Figure 4-4. Architecture of WheelCheX® as an example of an available WILD system

There are a variety of concepts to interpret wheel tread irregularities depending on the algorithm
and detection models of the available WILD systems. Two examples of wheel irregularity
conceptual interpretation are illustrated in Figure 4-5 and Figure 4-6.

49
 Figure 4-5. Voestalpine’s WILD system using optic sensors underneath of the rails [10]

Figure 4-6. An example of wheel tread irregularity detection concept of MERMEC’s WILD
system using accelerometers and strain gages [9]

4.1.4 Accuracy and Reliability of WILD System

The accuracy and reliability of a WILD system depends on two main factors:
• Technical characteristics of the system, including the type and algorithm of impact load
detector and sensors
• The number of WILD systems integrated that can verify the information of other WILD
systems

50
According to AAR’s Field Manual Appendix F-B.1(a) the following criteria should be met for a
functional WILD system [1, 29]:
“Wheel impact load detectors must be maintained such that each rail has at least 70% of
vertical circuits active. If less than 70% of the circuits are active on a rail, then the data
from that rail does not meet the validation requirements. The average vertical weight for
all wheels measured must be calculated for each active circuit. The range (maximum-
minimum) of these average weights for a rail must be less than 15 kips for any train set
containing 50 or more axles. If the range is greater than 15 kips, then data from that rail
does not meet the validation requirements.”
Table 4-2. WILD system thresholds to maintain the reliability and precision level of the
collected data
WILD System Resolution Speed Temperature Sensor Required Weight Ref.
Limits Technology Zone Accuracy
Ansaldo STS 11 lb. > 15 (-4 °F)–140 °F Fiber optical NA
2% [4]
mph clamps on rails
DeltaRail 1
NA NA NA Strain gages on NA
10% [67, 68]
the rail
L.B. Foster 2
100 lb. 30–180 32 °F–131 °F Strain gages on 50 feet
NA [59]
mph (Electronics) the rail
MERMEC NA 25–125 (-22 °F)–167 °F Accelerometers 35 feet
2% [9]
mph and strain gages
Progressive NA NA NA Strain gages on NA
Rail the rails 2% [63]
Technologies3
Schenck 0.2 mm 6–155 (-40 °F)–158 °F Strain gages on Approx.
Process wheel flat mph (Mechanics) concrete ties 15 feet
2% [61, 64]
depth 40 °F–86 °F
(Electronics)
Track IQ 5 mm spall 18–155 arctic, tropical Accelerometers 20–30
detection mph and desert and strain gages feet 3% [62, 77]
environment
Voestalpine NA > 12 (-22 °F)–158 °F Fiber optical Approx.
3% [10, 66]
mph clamps on rails 20 feet
1:
Currently managed by Vortok, part of Pandrol
2
: Including Nagory FOSTER system [80]
3
: Currently, part of “International Engineering and Global Connections”
NA: not available

In addition, the accuracy of WILD systems can be affected if the system is not properly operated
within the thresholds recommended by the respective manufacturers. These thresholds may
include the minimum and maximum speed of rail vehicles passing the WILD detector, the
environmental temperature of the WILD location, the stability of track structure, etc. Table 4-2
summarizes some of the characteristics and technical thresholds of different WILD systems (or
similar products) that can affect the system reliability and demonstrates the accuracy level of
different WILD systems available in the market.
In addition, other parameters may cause variation in readings from one pass over the WILD
system to the next due to [77]:
• Progression in an existing or a new irregularity created after the first reading

51
 • A narrow wheel-rail contact area
• Sharp edged, very small irregularities near the edge of the normal wheel-rail contact area
• Irregularities with a ‘resonant’ speed, usually well over 50 mph, in a way that wheel-rail
contact may become momentarily zero. Thus, the collected data can be affected by the
way the wheel ‘lands.’
• When a vehicle is turned around or a train reverses over a WILD array, it may aggravate
the irregularities; particularly it may increase the asymmetric defects.

4.1.5 WILD Database Management

4.1.5.1 Data Processing

Generally, the data collected by WILD wayside detectors can be processed using one of three
different approaches:
• On-site processing
• Remote (central office) processing
• Combined approach
The on-site processing approach relies only on the diagnostic and interpretational models located
in the control room next to the instrumented track. The remote processing approach uses a
computer model and scripts that are managed from a distance, typically located at a central
control office, to manage the data from all WILD and even other wayside detectors. The
combined approach takes advantage of both models. It may generate some basic reports and
temporary results using the on-site processing method, followed by more comprehensive and
advanced reports using remote (central office) computational tools. For instance, L.B. Foster, a
manufacturer of WILD systems, uses a combined processing approach. First it deploys
microcomputers at a control room “to sample the analog strain gage amplifier signal output and
determine the peak wheel load from each passing wheel,” Then a master computer is remotely
used for post-train passage analysis [77].
It should be noted that AAR’s InteRRIS® database system is considered a combined data
processing system in which all raw and processed data of individual WILD detectors are
warehoused under InteRRIS® to be further analyzed and studied by respective railroaders.

4.1.5.2 Data Storage

Each time a train passes the WILD site, a large amount of data is collected by the detectors. For
instance, assuming a 150-car freight train (600 axles) is passing along the WILD detector at 25
mph, which then requires continuous sampling for 6–8 minutes. For such a train, there will be
approximately 100 million data records collected by an eight-circuit WILD system [77].
The collected data can be classified by the WILD system at typically three levels:
• Train level: Particularly for transit and commuter rail systems with a pre-determined
train consist

52
 • Vehicle level: For all types of freight, passenger, and transit cars if they are equipped
with identification tags using RFID, AEI or other similar technologies
• Component level: Including wheel, axle, and truck components depending on the WILD
system capabilities
Typically, the following information is collected and reported by the WILD system [4, 7, 74]:
• Date and time of collected data
• Track number (in case of operating WILD systems on multiple-track corridors)
• Train direction
• Train speed
• Total number of axles
• Identifying the respective wheel, axle, and truck for collected data
• Gross weight of wheel, axle, truck, car, and train
• Dynamic Ratio for wheel, axle, truck, car, and train (average and max.)
• Total train weight
• Total train length
The stored WILD data should match with other available vehicle and equipment database data
such as the vehicle AEI tag (RFID), AAR’s UMLER® database (for verifying the vehicle
characteristics such as axle counts, car type, and wheel size) to be used by rail operators [44].

4.1.6 Communication Method

The WILD system can communicate within its on-site components, and other remote parts, and
staff using different technologies such as GPS-based communications, optical fiber, radio, or
modem-based systems [7, 10]. All these protocols and communication methods should be
reliable and integrated through the entire system. Thus, some of the WILD systems, including
products by Ansaldo STS [4] and L.B. Foster [59], were developed to include an auto-diagnostic
(self-diagnostic) feature. This feature can routinely check the operating conditions of the
individual components of the WILD system and automatically detect and report sensor failures
as well as measure equipment failures, including on-site processing and back office system
failures [4].

4.2 Site Selection Criteria of WILD System

As explained earlier in Section 2, there are several criteria and features for installing a new
wayside detector system, including for WILD systems, which should be considered when
studying the location of new system. This section briefly reviews the primary criteria regarding
new WILD system selection and system installation.

53
4.2.1 Track Characteristics
Depending on the type of WILD system and the manufacturer recommendations, there are
certain track characteristics that are critical when selecting the location of a new system, as
follows:
• Track Components: Most WILD systems can be installed on ballasted or non-ballasted
tracks, either concrete or wooden ties, though some systems are recommended to be
installed on concrete ties.
• Track Geometry Requirements: All WILD systems are recommended to be installed
along a tangent track with no vertical or horizontal curvature immediately located before
or after the WILD detectors.
• Track Structure Requirements: The WILD system may have location requirements
specified by the manufacturer, such as the system may not be located on a bridge on
wooden ties (or even slab track), along a grade-crossing, or within the boundaries of
switches.
• Ride Quality: A new WILD system may not be installed along a location where it can
aggravate the vibration and noise levels of train passage, particularly in proximity of
urban areas.

Figure 4-7. Comparison of impact loads on wood and concrete ties at 40 mph, 100-ton cars,
and machined flats [69] (recreated chart)
The characteristics of track components can affect the impact loads recorded by a WILD system.
For instance, tracks with concrete ties generally have higher stiffness in comparison to track with
wood ties. Thus, according to a study conducted by AAR [30], Figure 4-7, track with concrete
ties may generate 15–30 percent higher impact loads in comparison to wood tie tracks, at a given
speed and for a given flat spot.

54
4.2.2 Traffic and Speed Coverage
As explained earlier, the location of a WILD system should be chosen to cover as much traffic as
possible. Also, the operational speed of the rail vehicles passing along the given location should
comply with the speed threshold of the given WILD system recommended by the manufacturer.
Studies show that a WILD system can provide better coverage and more precise readings when
the system is installed along a track location with the operational speed of the trains in the range
of 40 to 50 mph [66, 78].
Note that the measured impact load is directly related to the rail vehicle speed for the range of
North American train operating speeds. In other words, the faster trains pass by the detectors, the
higher the wheel impact loads will be most likely generated for majority of wheel tread
irregularities. For instance, a review of WILD data for a wheel that was involved in a 2013
derailment on CP revealed that seven of the nine impacts detected by the WILD system for the
given wheel were measured at speeds between 30 and 42 mph, which were below the
recommended speed threshold of 50 mph of CP railroad for measuring the wheel impacts [78].
This underestimated the severity of the wheel irregularity. However, CP adjusts all measured
impacts to 50 mph. Table 4-3 presents the relationship between measured and calculated wheel
impact loads at different speed levels conducted by CP.
Table 4-3. An example of measured and calculated wheel impacts at various speeds
(source: CP) [78]
Train Nominal Measured Calculated wheel
speed weight per wheel impact impact at 50 mph
(mph) wheel (kips) (kips) (kips)
30 33 90 128
35 33 90 114
40 33 90 104
50 33 90 90

In another study conducted by the AAR, several wheels with a given level of wheel irregularities
(mainly flat and shell spots) were operated over a WILD detector to evaluate the impact of speed
on the WILD readings. The results of this study confirm that a higher speed may generate a
higher peak load (impact load) readings by the WILD system for the same wheel irregularity
compared to lower speeds as shown in Figure 4-8. The study also showed that higher speeds
(e.g., 70 mph) lead to a more scattered range of peak loads and mean load values, compared to
lower speeds (e.g., 20 mph), resulting in a higher standard deviation in the WILD readings from
different WILD detectors for the same wheels, Figure 4-8 and Figure 4-9.

55
Figure 4-8. Impact load scatter (36-inch diameter out-of-round wheel, single 140 mm divot,
100-ton loaded car [69] (recreated chart)

Figure 4-9. Impact load statistics (36-inch diameter out-of-round wheel, single 140 mm
divot, 100-ton loaded car [69] (recreated chart)

56
4.2.3 Location Accessibility and Maintenance Features
A new WILD site should be accessible for internal inspection and maintenance activities on the
system components. In addition, there should be a rail facility located within a reasonable
distance from the proposed location of the WILD site, such that train crew can stop a train to
inspect or set out cars flagged by the system, in case a wheel with very high peak loads (or high
DRs) was detected, requiring immediate attention.
Also, it is typically recommended to install a WILD system outside of train stop locations (e.g.,
away from yards, sidings, stations, or hump yard boundaries), which tend to create high
variability in operating speeds.
The WILD system may be integrated with other existing wayside detectors such as AEI reader,
HBD, DED, THD, and HWD. However, the manufacturer of WILD system may not recommend
installation of the WILD system within certain proximity of other existing wayside detectors to
avoid any interference with them, or other rail equipment and facilities.

4.3 WILD System Thresholds, and Performance Monitoring

4.3.1 Data Triggers and Thresholds

The data collected from the WILD system should be analyzed for any further actionable or
informative requirements, either automatically by the detector software or by the respective
authorized parties (rail operator, car owner, etc.). The analysis should be reported in near real
time for those wheel irregularities that can endanger the safety of rail vehicle operations when a
high impact load is detected by the WILD system.
The AAR has developed a Field Manual of the AAR Interchange Rules that can be used for
analyzing the WILD system collected data. According to Rule 41, the InteRRIS® database has
four performance indices for WILD systems to flag a wheel for action [7]. The performance
indices of InteRRIS® database for WILD detector systems are shown in Table 4-4.
Table 4-4. WILD indices for freight service [7]

Index* Description
The operating railroad is required to inspect the train and move
Final Alert (140 Kips and
it at slow speed (below 30 mph) to set out the affected car for
above)
repair.
AAR Condemnable (90–140 The operating railroad is required to shop the car for repair as
Kips) soon as the car reaches destination.
If a car is shopped for any non-wheel related repairs, the repair
Opportunistic Repair (80–90 facility is permitted to make wheel repairs to eliminate defects
Kips) causing high dynamic loads and recover the costs from the car
owner under AAR’s Interchange Car Repair Billing system.
The car owner can choose to shop the car and make wheel
Window Open (65–80 Kips)
repairs to eliminate defects causing high dynamic loads.
*
Association of American Railroads - Field Manual Rule 41(r)

The performance indices shown in Table 4-4 are more applicable for freight traffic since freight
rail cars have higher axle loads (and subsequently higher impact loads) than passenger rail

57
vehicles. Thus, similar performance indices can be defined and used for passenger and rail transit
industry based on incorporating the impact load factor (or DR) 4 [1].
Freight railroads may also use the concept of DR (used by passenger and transit services) to
capture empty or lightly loaded rail vehicles with defective wheels. In addition to the actionable
indices of impact loads recommended by the AAR (90 and 140 kips loads), freight service may
also use the following indices based on the concept of dynamic load parameter [1]:
• Dynamic Load increment (over the static load) ≥ 30 kips or
• Dynamic Ratio (Dynamic wheel load/static wheel load) ≥ 3.0 or
• Peak Impact (dynamic wheel load) ≥ 65 kips
Passenger and transit rail service operators typically use the concept of DR due to the small
deviation between empty and loaded weights of passenger cars. Table 4-5 presents the actionable
thresholds and triggers for WILD systems based on DR criteria that was designed for and
operated by MNR commuter railroad.
Table 4-5. MNR’s decision table for “Actionable” WILD data [1]
Threshold Level Dynamic Ratio (MNR Criteria) Action
Emergency ≥ 5.00 Defective wheels addressed
4.00–4.99 immediately
(Loco Wheel: 136,000–170,000 lbs.)
Alarm (Coach Wheel: 71,000–89,000 lbs.)
3.00–3.99 Scheduled for timely correction
(Loco Wheel: 102,000–136,000 lbs.)
Warning (Coach Wheel: 53,4000–71,000 lbs.)
Information 2.00–2.99 Addressed during normal
maintenance
No Notification 1.00–1.99 Information only (Normal
operation)

Any other thresholds or warnings set up by rail operators or vehicle owners should be developed
and applied in a way that all relevant condemnable thresholds or defect sizes recommended by
the AAR, or prescribed in the FRA regulations (or Transport Canada, in case of operating trains
through Canadian territory), are met [44]. Thresholds for a new WILD system should be
established in consideration with the maintenance shop capacity 5 especially for corridors that
have limited shop capacity, such as urban transit or commuter rail systems. Temporary
thresholds should be tested for a short period of time to evaluate the ability of the shops to true
the number of flagged wheels triggered by the temporary thresholds of the new WILD system.

4.3.2 Decision Flow and Action Assignment

System integration is part of a successful implementation for any new detector system being
installed. In the case of a new WILD detector, the system integration plan should be developed
jointly by the departments and entities who are involved in the daily operations, such as:

4
Dynamic Ratio: The ratio between the peak dynamic load and the measured static load for a given wheel
5
: Number of cars which can be concurrently accommodated at the shops.

58
 • Information technology and operation management database center
• Track and rolling stock maintenance and inspection planning
• Train control and signaling systems, particularly for any commendable action identified
by the detectors
According to Rule 41 of the AAR Field Manual, the following actions should be considered
when determining steel wheel defects:
1. Condemnable at Any Time:
Wheel Out-of-Round or 90,000 Pounds (90 kips) or Greater Impact.
(1) Detected by a wheel impact load detector reading 90,000 pounds (90 kips) or greater
for a single wheel. The detector used must meet the calibration and validation
requirements of Appendix F. The detector must reliably measure peak impacts and must
provide a printable record of such measurements. Device calibration records must be
maintained. Wheels with condemnable slid flat spot(s) 6 are handling line responsibility
and must not be billed otherwise.
2. Condemnable When Car Is in Shop or Repair Track for Any Reason
Detected by a Wheel Impact Load Detector reading from 80 kips to less than 90 kips for
a single wheel. The detector used must have been calibrated per Appendix F. The
detector must reliably measure peak impact and must provide a printable record of such
measurements. Device calibration records must be maintained. Wheels with
condemnable slid flat spots are handling line responsibility and must not be billed
otherwise. This will be considered an Opportunistic Repair for the repairing party.
Additional examples of decision flow and action assignments are explained in following part
(Section 4.5), based on experiences reported by CN, CP and MNR railroads.

4.3.3 WILD Data Monitoring

Monitoring the status of the wheels with sub critical irregularities can prevent unexpected and
dramatic failures in the future, especially if a growth trend is identified by analyzing the time
history of maximum wheel loads. The car owners can monitor wheels under empty car
conditions as well as using other criteria, such as DR and dynamic increment. For instance,
according to AAR Rule 41 for freight rail operations, a DR greater than or equal to 3.0 and a
dynamic increment greater than or equal to 30,000 lbs. can be flagged for car inspection and
repair [1].
The procedure for monitoring the status of the wheels may or may not be automatically handled
by the WILD systems available in the market. Thus, additional models and manually defined
algorithms might be needed to frequently monitor the impact load trend of queried cars. This
section briefly reviews some primary aspects of WILD data monitoring that can be applied in
both freight and passenger rail operations. More examples of WILD data monitoring conducted
by rail industry are presented in Section 4.5.

6
Slid Flat:
a. 2 inches or over in length.
b. 2 or more adjoining spots each 1½ inch or over in length.

59
4.3.3.1 System Monitoring on WILD Assigned Actions
Rail operators and car owners or shippers may have a different Asset Management System
(AMS) or Inspection Management System (IMS) to monitor their rolling stock health status
before and after applying any maintenance or inspection activities.
As previously mentioned, wheels with high impact loads can be again flagged in future WILD
measurements with even higher impact loads or more serious wheel irregularities, even if they
have been set out for additional inspection or repair activities [7, 78]. Thus, monitoring the health
status of a flagged wheelset after conducting the proper action can be beneficial to develop a
trend-based process wherein repair histories of fleet vehicles can be reviewed for further
inspection or maintenance activities. Such a monitoring system can help rail maintenance crews
ensure that the remedial action corresponding to the WILD alerts was effective in correcting the
problem.

4.3.3.2 Monitoring Seasonal Patterns

Monitoring the data collected by WILD detectors can help railroad operators and maintenance
departments investigate high wheel impact loads caused by seasonal patterns.
As shown in Figure 4-10, since implementation of WILD systems across North America’s rail
network, the incidences of high wheel impacts have decreased, while both rail traffic and the
number of WILD detectors have gradually increased. Also, the diagram demonstrates that overall
there are more wheel impacts observed during the winter season (December, January, February,
and March) of each year compared to the rest of the year. However, the overall incidences of
winter peak loads were reduced from 7 per 1,000 to 2 per 1,000 wheels inspected over the 2003–
2013 period [8].
Another study conducted in Finland also confirmed that the incidences of high wheel impacts
increased during winter months (December–February) in comparison to the rest of the year [81].

60
 Figure 4-10. Rate of WILD readings (per 1,000 wheels) and seasonal effect in North
America during 2003–2013 [8]

4.3.3.3 Sharing WILD Data with Other Parties and Rail Authorities
WILD data must be shared between the owner of the track operated on and any other railroad
operating over this track. Sharing WILD information is beneficial for both parties on account of:
• The host railroad can review and integrate the history of WILD database of other
operators’ cars with its own WILD database, such as AAR’s InteRRIS®, to pay extra
attention to those cars that have been previously flagged with certain impact loads.
• If the host railroad captures any high impacts on another operator’s rail car, sharing such
information through an integrated WILD database system can facilitate and improve the
future wheelset maintenance, repair and inspection activities.

4.3.3.4 WILD Alerts Closure

As described in previous sections, generated alerts and condemnable actions should be
monitored for proper responses to be taken. Following appropriate action by the respective
authorities or maintenance crews, the alerts should be closed and archived for future reference.
For instance, the data summaries developed by L.B. Foster can either be closed manually when
an appropriate repair or inspection is reported, or they can be closed automatically based on the
criteria presented in Exhibit 69 of L.B. Foster’s manual [50].
In the case of integrating the WILD system under the EHMS platform, it should be noted that the
alert data and the alert state (open or closed) are always available to all EHMS users. For all
other raw and processed data, EHMS utilizes UMLER®, a tool integrated under the EHMS
system, to display the details of requested information after validating the ownership and
maintenance status. Similarly, the details on alert closure information are also restricted, but the
authorities who are performing or reporting the condemnable action always have access to their
closure reports [50].

61
If an action (repair, replacements, inspection, etc.) was not correctly reported, the closure can be
deleted by a qualified and authorized party defined in the system. It should be noted that deleting
a closure will re-open the reported alert unless another repair action would address the issue,
which closes the alert [50].

4.3.4 WILD Data Trending and Advanced Analysis

In addition to monitoring the WILD data for improved efficiency levels for tracking the wheel
irregularities and monitoring the health status of the rail fleet, certain long-term trends can be
also extracted from studying the WILD data. For instance, some studies show that skids (or flats)
can become worn at the edges, but then can sharpen again at the next heavy braking application.
Such flat spots tend to skid more easily, and it is especially common on locomotives [77].
Another study [73] examined the balance between the impact loads of a wheel pair on the same
axle, which can be highlighted as one of the long-term trending aspects of a WILD system. The
study focused on the issue that typically failed wheels and their mate wheels have a significant
difference between their impact load readings (approximately 2,400 wheelsets per day in North
America). Thus, this study hypothesized that removing wheels with a high difference in their
load readings may capture failed wheels. However, removing such wheels may not be the most
cost-efficient alternative [73], because another study demonstrated that approximately 12 percent
of high loads were caused due to imbalance loads on the left and right wheels, and not related to
the wheel irregularities [78].
This section briefly reviews some of the long-term trends of analyzing the collected WILD data
that are indirectly related to the wheel irregularities. Capturing such issues can eventually reduce
the operating disturbances, financial costs, and environmental damages of railroads.
Similar to the previous section, more examples of how railroads conduct WILD data trending are
provided below.

4.3.4.1 Impact of WILD System on Track Related Derailment Reduction

Rolling stock (equipment) and track-related irregularities are two major causes of train
derailment accidents. The rolling stock related issues can be typically detected using other
technologies of wayside detector systems such as HBD, TPD, and THD, rather than WILD
system. However, there are more opportunities to prevent track-related derailments by using
WILD systems. To analyze the long-term trend of applying WILD for track-related derailments,
queries were collected and classified for the WILD systems that have been in service throughout
the national freight rail network during the period 2003–2012.
Figure 4-11 presents the overall trend of the number of WILD systems installed during 2003–
2012, compared to the annual trend of track-related derailments (including top 10 and all causes
combined). As shown in Figure 4-11 the number of WILD systems has grown from 53 in 2003 to
169 in 2012 (219 percent growth), while the number of track-related derailments has reduced
from 154 in 2003 to 72 in 2012 (reduction of 53 percent). The overall trend for the top 10 track-
related causes is similar and shows a reduction in derailments from 93 to 31; a reduction of 66
percent. This comparison indicates that the WILD system decreased the number of track-related
derailments during the period of 2003–2012 data collection [7], with the potential to further
reduce derailments as more WILD sites come on-line.

62
 Figure 4-11. Impact of WILD installations on track cause related derailments [7]
By evaluating the WILD data on a given wheel, it is recommended to inspect the given wheel for
any further proper action (removal, speed reduction, etc.) if there is any trend in increasing the
impact loads for the given wheel. The reason behind such a recommendation is that there may be
a correlation between elevated WILD data and wheel or track related irregularities in a way that
can cause serious rail accidents including train derailments [78].

4.3.4.2 WILD System Performance vs. Wheel Maintenance Actions

Another aspect of WILD system trending would be a comparative analysis between flagged
wheels captured by WILD detectors versus the status of inspection and maintenance records
applied on the same wheels in the shops. Such comparative analysis could evaluate the
performance of a WILD system in terms of any correlation between WILD detection status and
respective maintenance actions applied on the same wheels in the shops. Generally, Receiver
Operating Characteristics (ROC) methodology can be used to justify the functionality of a given
WILD system. Four primary conditions of applying ROC method between WILD detection and
maintenance actions on the same wheels can be hypothesized as below:
• True Negative: A wheel on a given car “was not flagged by WILD,” and the car “was
not repaired in the shop” within a reasonable time horizon.
• True Positive: A wheel on a given car “was flagged by WILD,” and the car “was
“repaired in the shop” within a reasonable time horizon.
• False Positive: A wheel on a given car “was flagged by WILD,” but the car “was not
repaired in the shop” within a reasonable time horizon (known as Type I error).
• False Negative: A wheel on a given car “was not flagged by WILD,” but the car was
“repaired in the shop” for causes which could have been detected by WILD system
(known as Type II error).

63
Note that a properly functioning WILD system should not have a relatively high number of false
positive or false negative records.

4.3.4.3 Analyzing WILD Performance across Multiple Locations

It is very common to operate WILD detectors on multiple tracks at a particular WILD site, or
across different locations in a way that the data can be compared and analyzed if common
locomotives and cars pass the sites.
Although several factors can affect the performance of a WILD system (including number of
wheel passages, speed, etc.), comparison between different WILD detectors along a given
network can help evaluate the performance of each WILD system, especially if the passing
traffic is relatively similar along the given WILD systems (e.g., a commuter rail system or cycle
freight trains).
More details of evaluating different WILD detectors are explained in Section 4.5 of this guide
based on MNR’s WILD data analysis.

4.4 Calibration and Maintenance Requirements

4.4.1 Calibration Method and Frequency

According to the AAR, WILD systems must be calibrated at least once in a 3-year period [1].
However, manufacturers typically recommend calibration more frequently, which could be every
6 months or once a year [12].
The calibration procedure is typically carried out by using a rail vehicle with a well-known
weight such as a locomotive or an empty or fully loaded rail car, making several passes over the
site. Vehicles with weights as widely spaced as possible (such as a locomotive and empty car)
should be chosen for the calibration procedure. This wide load range will assist with verifying
the linearity of the system. Furthermore, these vehicles should be periodically operated over the
WILD site(s) to verify that the calibration of the system has not drifted since the previous
scheduled calibration [1]. For instance, as recommended in the WheelCheX® user instruction
manual for MNR calibration purposes, at least 1,000 axles should have passed over each track
before the bays of track sensors are equalized [12]. In the case of using non-locomotive vehicles,
the empty freight cars or late evening or early morning commuter trains may be used for
calibration purposes since they are likely to be lightly loaded or even empty providing a
relatively consistent vertical load.
Also, it is recommended the calibration passes be conducted at the same range of operational
speed of the daily trains that typically traverse the WILD site.
When a new WILD site is calibrated for the first time, or when calibrating a WILD site following
the replacement of any track sensors, it is important to perform the system calibration with the
default calibration values as recommended by the manufacturer [12].
It should be mentioned that some of the WILD systems, including ATLAS FO (by Voestalpine),
MultiRail® WheelScan (by Schenck Process) and WILD (by MERMEC, and Ansaldo STS), may
not require running a particular rail vehicle for calibration procedure as these systems claim to be
automatically calibrated [4, 9, 10] or provide a high precision level before static calibration [61].

64
4.4.2 Calibration and Maintenance Procedures
The AAR Field Manual-Appendix F (page 742), explains the calibration and functional
requirements for different wayside detector systems which are installed in the freight rail
network. According to this document, the wayside detectors, including WILD system, should be
calibrated using the procedure specified by the OEM as described below [1]:
• Static calibration should be applied based on the manufacturer's procedures at installation
and, at a minimum, once every 3 years thereafter.
• The calibration procedure should be stored with the calibration record.
• Calibration records should be available upon request.
For instance, MNR’s WILD system, WheelCheX®, is calibrated and maintained based on the
instructions specified by DeltaRail (the system manufacturer) and managed by MNR’s
Communication and Signal (C&S) department. The C&S crews perform maintenance of the
WILD system twice a year and apply the system calibration once a year, which is more frequent
than the AAR requirement of once every 3 years. The higher maintenance frequency conducted
by the C&S crews of MNR is likely due to more traffic passing over the track and due to the
sensitivity to noise.
During maintenance events, MNR remotely collects WILD data for known vehicles, which are
used for a system check versus calibration values. The calibration procedure is applied while
only one of the four tracks at MNR’s WILD site is out of service. Based on the type of
calibration method, the measurements of the MNR’s WILD system can be within ±5% or ±10%
of a typical locomotive axle load, which meets or exceeds the claimed accuracy of the DeltaRail
WILD system shown in Table 4-2 [1].
As mentioned earlier, the calibration procedure of a given wayside detector may need to be
facilitated by additional programing, macros, or specific tools that the WILD manufacturer has
provided in advance. For instance, according to the instructions provided for MNR’s WILD
system, the calibration process of WheelCheX® uses certain macros embedded in a spreadsheet
containing specific analysis instructions [12].

4.4.3 Use of Collected WILD Data for Improving Calibration Procedure

As discussed earlier, analyzing different aspects of a wayside detector system including WILD
can provide data for determining whether a re-calibration or a maintenance action is required for
the WILD components or not. For instance, the collected mean loads of WILD data on a given
rail vehicle can be assessed through a long-term analytical approach, to verify the calibration
procedure. If the collected data on the given car are not within the expected range of mean load
values, then the calibration procedure may need to be investigated. This could be indicative of
strain gage drift. More details regarding this subject are explained in Section 4.5.

4.5 Examples of Current WILD Experiences and Best Practices

This section briefly summarizes experiences of selected railroads as best practices regarding
operating WILD systems in their networks. The materials covered in this section include
examples from both freight railroads (such as AAR, CN, and CP), as well as passenger railroads
including MNR.

65
4.5.1 WILD System Action Assignment and Decision Flow

4.5.1.1 CN Experiences
Currently, CN railroad has developed the following WILD alarm thresholds for measured wheel
impact loads over 140 kips [78]:
• Cars with a single measured impact over 160 kips, or 200 kips for a calculated impact,
should be immediately restricted with a speed limit of 25 mph. If the train is operated in
the inbound direction, the car must be set out at the terminal, but if the train is an
outbound train, the car must be set out at the first designated siding. In addition, the car
will be bad ordered by a qualified mechanical crew for required repair and inspection
services.
• Cars with a single measured impact between 150 and 159 kips should be immediately
restricted to 10 mph less than the speed recorded at the WILD site. The Rail Traffic
Controller (RTC) will then decide whether the car should be set out at the inbound
terminal (if inbound) or at the first designated set-out location (if outbound). If the set-out
location is not available, the car can proceed up to another convenient location for set-out
but cannot move beyond the next set-out location, where a certified car inspection (CCI)
will be issued for the car. The car will be bad-ordered by a qualified mechanical crew for
required repair and inspection services.
• Cars with a single measured impact between 140 and 149 kips should be immediately
restricted to 5 mph less than the speed recorded at the WILD site. If the temperature at
the WILD site is -13 °F (-25 °C) or colder, the speed reduction must be 10 mph less than
the speed recorded at the WILD site. The RTC will then decide whether the car should be
set out at the inbound terminal (if inbound) or at the first designated set-out location (if
outbound). If the set-out location is not available, the car can proceed up to another
convenient location for set-out but cannot move beyond the next location, where a CCI
will be issued for the car. The car will be bad ordered by a qualified mechanical crew for
required repair and inspection services.
• In all above-mentioned cases, the flagged wheel should be replaced before returning the
car to regular service.
In addition, CN has developed the following guidelines for handling cars with measured (peak)
impact loads between 80 and 139 kips [78]:
• Cars arriving from interchange to CN with wheel impacts are automatically identified
• Wheel set removal between 80 and 89 kips when a car is on a shop or repair track
• Automatically identify wheels with impacts between 90 and 139 kips
• Wheel impact between 90 and 139 kips is removed selectively according to AAR
guidelines at CCI locations

4.5.1.2 CP Experiences
Similar to CN regulations on WILD action assignments and thresholds, CP requires the
following thresholds and action assignments as a WILD alert is generated [78]:

66
 • In the northern Ontario region, a car should be bad-ordered immediately if the wheel
impact is measured higher than 130 kips, or the wheel impact adjusted to 50 mph is over
150 kips. For the rest of the CP network, these thresholds are adjusted to 140 kips or 170
kips, respectively, for measured and adjusted wheel impacts. Following the action
assignment, the train speed should be reduced, and the flagged car should be set out at the
next designated location for required repair and replacement action on the flagged
wheelset.
• For adjusted impact loads higher than 90 kips, the flagged car should be bad-ordered as
soon as it is empty. Such a requirement allows the car to be normally operated up to its
destination and then it can be repaired once it is unloaded. In addition, CP has developed
several opportunistic threshold limits for the cars with adjusted impact loads between 90
and 110 kip. In these cases, the car is flagged in CP’s car information management
system, but similar to the previous situation, it will not be bad-ordered until arrives at its
destination without any restrictions. Then, the car can be repaired (with or without
removing the flagged wheelset) when operationally convenient.

4.5.1.3 MNR Experiences

MNR, as one of premier commuter rail services in North America, has successfully integrated
the operations of their new WILD system across multiple departments via the following [1]:
• The Signals department has implemented and operates the WILD installation
• The Operations department monitors alerts and routes the actions resulting from the
alarms
• The Mechanical department handles the resulting inspections and maintenance
• The Information Technology (IT) department has developed, implemented and maintains
the email communication chains as well as the data storage
After the data is collected from MNR’s WILD detectors, it is processed by the trackside
computer to generate a report for every train passing over the WILD site. Then the processed
report is transmitted to MNR’s IT group to process, store, and merge with the train database.
In case of an “Alert” message, it is then communicated to MNR’s Operations Control Center
(OCC) for follow-up action, including inspection, repair and maintenance activities that are
managed by MNR’s Mechanical department. Figure 4-12 provides an overview of the
interdepartmental communications that are carried out at MNR during day-to-day operations of
the WILD system [1].

67
 Figure 4-12. Interdepartmental data flow for MNR's WILD system [1]
At MNR’s OCC department, the RTC monitors all trains operating along the MNR network
including foreign trains (Amtrak and freight trains). The monitoring by RTC crews includes
solving various issues that trains may encounter en-route including the notifications that they
receive from wayside detectors such as WILD systems [1].
When an actionable alert is generated by a WILD system, several communication events may
occur as follows [1]:
1. Automatic email messages are sent to various departments, including Operations and
Mechanical
2. The alerts are broadcast on an MNR internal web page, (see Appendix B) for view by a
larger audience
3. The alerts are broadcast to a dedicated screen at the OCC
4. An IMS/AMS shop order is created to track the incident and the required activity
It should be mentioned that any wheel irregularities identified by MNR’s WILD system is
typically reported within 2 hours after detection, based on different levels of WILD thresholds
[68].
The follow-up actions are assigned depending on the severity of the alert. Figure 4-13 depicts
MNR’s actions following a WILD alert generated for a train passing by the detectors. Also, the
action and information sequence following generating WILD alert is shown in Figure 4-14 [1].

68
 Figure 4-13. Follow-up actions for a WILD alert at MNR [1]
As shown in Figure 4-13 and Figure 4-14, the action assignment following a WILD alert can be
summarized as follows [1]:
• In case of an “Emergency” alert (Dynamic Ratio ≥5.0), an e-mail is sent to OPS, FMO
and MofE, to schedule a wheelset replacement by Qualified Maintenance Personnel
(QMP) at the next available site
• In case of an “Alarm” alert (5> Dynamic Ratio ≥4.0), an e-mail is sent to OPS, FMO and
MofE, to schedule a wheelset inspection by QMP at the next available site
• In case of a “Warning” alert (4> Dynamic Ratio ≥3.0), an e-mail is sent to FMO and
MofE to schedule an inspection before Calendar Day Mechanical Inspection (CDMI)
• In case of an “Information” alert (3> Dynamic Ratio ≥2.0), the QMP should inspect the
flagged wheelset at the next scheduled CDMI with special attention
These actions are defined by MNR’s Mechanical Department through MMA#292 forms as
shown in Appendix C

69
 Figure 4-14. Information/action sequence of WILD system at MNR [1]

4.5.2 Monitoring WILD Assigned Actions

4.5.2.1 Freight Railroad Experiences

Several analyses were conducted on WILD data collected on freight trains to monitor the WILD
alerts and assigned actions. One example of such practice would be monitoring the time history
of maximum wheel loads captured by all WILD systems that a given rail vehicle passed over. As
shown in Figure 4-15, the captured loads correspond to all eight wheels on a given freight car
including both loaded and empty car conditions. The wheel loads that result in removal due to
“AAR Condemnable” or “Final Alert “conditions are shown in Figure 4-15 for loaded car
conditions [7].
As depicted in Figure 4-15, the queried car had wheels removed three times due to exceeding
AAR Condemnable criteria and once due to a Final Alert. The first wheel removal event was
triggered by the AAR Condemnable limit on wheels R1 and R2 (right wheels on B-end axles) on
April 15, 2005. These wheels ran with no high impact loads until April 2010 when a high impact
load exceeding AAR Condemnable level was identified on wheel R2. The car was also flagged
on November 11, 2008, when a Final Alert level was exceeded on wheel L4 with no high impact

70
load history. The same wheel was detected 3.5 years later on March 26, 2012, where it was
identified for a vertical load exceeding the AAR Condemnable limit of 90 kips [7].

Figure 4-15. An example of maximum loads time history of a 100-ton 4-axle car wheels—
empty and loaded conditions during 2004–2012 [7]
Note that, as shown in Figure 4-15, wheel R2 had been initially flagged on the same day (May
23, 2010) approximately 21 hours before the removal action occurred due to exceeding the AAR
Condemnable criteria (greater than 90-kips level). Further investigation revealed that the only
difference between these two events within such a short time frame was that the train speeds
recorded by the WILD system were different. The first wheel detection event was while WILD
data was collected with the train passing the WILD site at a speed of 42.6 mph. The same wheel
passed the other WILD site at a speed of 48.7 mph at the next wheel detection event. Thus, the
higher speed (even if it was only approximately 6 mph higher than at the previous WILD site)
caused higher impact loads to be captured by the WILD detectors [7]. These observations
confirm that alert triggers at a WILD site are speed-dependent, as previously discussed.

4.5.2.2 MNR Experiences

MNR has an embedded model for monitoring WILD alerts through a web-based Graphic User
Interface (GUI), which is integrated within their AMS. Under this system, once a train is
selected, the WILD alerts associated with given train axles can be displayed to determine if there
are any warnings generated on the selected train (Figure 4-16). The users of this system are also
able to search the WILD history of a given car for 1 week periods for all trains the car was in.
This ability allows the user to determine what wheel load levels the car was experiencing before
reaching the respective WILD thresholds (Warning, Alarm or Emergency level) [1].

71
 Figure 4-16. An example of AMS GUI details for a selected MNR train [1]

4.5.3 Best Practices on Monitoring Seasonal Patterns of WILD Data

4.5.3.1 Freight Railroad Experiences

As part of a study conducted by SA on WILD records of freight trains during 2004–2012, there
was an analysis of the seasonal pattern of WILD alerts. Figure 4-17 shows the seasonal pattern of
WILD data extracted from the study. As depicted in Figure 4-17, a consistent pattern in the
number of higher impact loads is present during the winter months. The impact loads are
between 80–300 percent higher in the winter months than the rest of the year [7].
A research effort conducted by AAR’s Wheel Defect Prevention Research Consortium 7 program
shows that higher impact loads during the winter months are caused because generally wheels
tend to develop more shelling issues in the winter months, especially when trains are operating in
snow conditions. The reason is that under snow conditions, snow can melt at the contact patch
and infiltrate into cracks. When the defect point of wheel is moved away from the wheel-rail
contact area, the melted snow within the wheel crack may freeze due to exposure to the ambient
conditions, which can cause crack widening leading to a more rapid defect deterioration rate.
Repeating such conditions over time can accelerate the crack growth and resulting in a shelling
issue on the wheel tread [7].

7
Steven L. Dedmon et al., “Accelerated Shelling –A Winter-Time Problem,” Wheel Shelling/Failure Study Group,
June 2007, Topeka, KS.

72
 Figure 4-17. “AAR Condemnable” wheels for the US WILD systems (monthly count) [7]
Another study conducted in Australia also shows that wheel irregularities can not only cause
more rapidly deteriorating wheel bearing failures, but that they are also more frequent during the
winter months (May to September). Although all bearings were re-greased to reduce the failure
rate during the winter months, installing a new WILD system demonstrated that all failed
bearings had relatively higher wheel impact loads as well. Therefore, this result strongly suggests
that re-greasing the bearings did not address issues related to the bearings, but the increased
bearing failure rate in the colder months was mainly caused by wheel irregularities with a higher
deterioration rate [77].

4.5.3.2 MNR Experiences

Another example of a seasonal pattern reflected by WILD systems can be derived from MNR’s
WILD data, showing similar behavior for commuter rail service. As shown in Figure 4-18, a
seasonal trend of an increased number of flagged wheels can be observed during the fall season
(particularly November) along MNR network. The reason behind flagging more wheels by
MNR’s WILD system in the fall season is due to the fact that the majority of MNR tracks are
surrounded by trees along the MNR’s right of way, which drop leaves onto the track. The wheel-
rail friction coefficient is reduced when leaves are on the track. Thus, during the fall season there
is always a higher risk of experiencing flat spots caused by applying wheel brakes on the fallen
leaves on the rails. For instance, as shown in the figure, there are approximately 156.5 flagged
wheels captured during November, on average, which compared to the average rate of capturing
wheel irregularities during the rest of the year (19.5) is approximately eight times higher.

73
 Figure 4-18. Seasonal trend on number of flagged wheels by MNR’s WILD system during
the fall season (particularly November) of 2010–2015
(Note: the order of Nov and Dec columns were switched in the figure, for the sake of better illustration)

4.5.4 WILD System Performance vs. Wheel Maintenance Actions

As mentioned earlier, evaluating a WILD system by comparing the flagged wheels to wheel
maintenance records can demonstrate the accuracy of the given WILD system in terms of
minimizing false positive and false negative alerts. For instance, Figure 4-19 compares the
number of wheels flagged by WILD detectors for a selected car of MNR’s network versus the
number of truing actions which have been applied on this car in the shops due to only “flat and
shell spot” irregularities, during the 2010–2015 data collection period.
When comparing these two criteria, the detection/maintenance effectiveness ratio 8 was observed
with values of 8.5%, 9.5%, 14.6%, 5.4%, 6.7%, and 7.0% respectively for 2010–2015. Since the
“truing records” have been filtered for “only flat and shell spots,” which are typically detectable
by WILD systems, a strong correlation between these two factors (flagged by WILD and trued in
the shops) would be present. It is assumed that the WILD system is the only major
technology/method for detecting the wheel irregularities (mainly flat and shell spots), and so this
ratio can be used as a benchmark index to quickly evaluate the WILD system performance.

8
Effectiveness Ratio = “# of flagged wheels by WILD” divided to “# of truing actions due to flat & shell spots”

74
However, there are other parameters that should be considered when comparing these two
criteria:
• Some of the flat and shell spots could be captured and treated during preventive
inspections before they can be flagged and detected by WILD systems.
• Whenever a wheel is determined to be trued due to any high impact causes, the other
wheel on the same axle (and sometimes even other wheels on the same truck or on both
trucks) should be also trued due to other regulations (maintaining a similar diameter for
the two wheels on any one axle).
• There is always a chance that a wheel with a flat or shell spot may pass the WILD
detector without passing the affected area over the strain gages or sensors (depending on
the type of the WILD system, number of sensors or strain gages, distance between
sensors on the WILD site, wheel size, and wheel passage speed). Therefore, the chance of
wheel irregularity detection may vary on different WILD sites.
Thus, considering the above-mentioned criteria and assumptions, the respective “Effectiveness
Ratio” of the WILD system can be bounded within thresholds of 1 to 50 percent. A WILD
system with 50 percent effectiveness ratio can be interpreted as a system that captured all
respective wheel irregularities that have been trued, while the mating wheel of the defective
wheel was also trued to provide smooth dynamic behavior. Obviously, a high value of
effectiveness ratio for a given WILD system (from 25 percent up to 50 percent) means it is a
more effective system for capturing wheel irregularities in terms of flats, shell spots, and other
issues which can cause high impact loads.
By evaluating the effectiveness ratio of the selected car series presented in Figure 4-19, we can
observe that the given WILD system was more effective during 2012 with effectiveness ratio of
15 percent, in comparison to other years that the system effectiveness was under 10 percent
thresholds.

Figure 4-19. Comparison between number of flagged wheels by WILD and number of
truing due to “flat and shell spot” irregularities, collected for the given car series of MNR
during 2010–2015

75
To compare the WILD effectiveness with the actual maintenance actions applied on a given car
series, a particular car was randomly selected and tracked in terms of both WILD and
maintenance records (either truing or change-out actions) during 2010–2015. Figure 4-20 depicts
the histogram of all flagged wheels detected by WILD systems for the selected car, and Table
4-6 summarizes all maintenance actions (truing and change-out) applied on this car during the
2010–2015 data collection period.

Figure 4-20. Histogram of flagged wheels by WILD system for a given car of MNR
commuter RR during 2010-2015
As shown in Figure 4-20, there are two WILD events (A and B highlighted in orange) that were
triggered by the WILD system, but no maintenance action was recorded for the car in response to
these WILD events (false positive) as shown in Table 4-6. As shown in Table 4-6, each
maintenance action was classified as to whether it was triggered by the WILD system or was
conducted through preventative maintenance/inspection (first column of the table). According to
Table 4-6, four maintenance actions on the car were conducted due to WILD trigger system
alerts within a few days after detection, while three truing actions were performed on this car due
to inspection rather than being detected by the WILD system. Note that the irregularities that
resulted in the change-out performed on October 11, 2013, could not be detected by the WILD
system due to the nature of this wheel defect.

76
 Table 4-6. History of “Truing” and “Change-out” actions on the wheels of the same given
car of MNR commuter railroad during 2010–2015

Overall, we can conclude that the WILD system had a fair correlation with the maintenance
actions of the selected car as presented in Table 4-6 (three truing actions detected by WILD out
of six detectable truing records).

4.5.5 Improving Calibration Procedure Using Collected WILD Data

As explained earlier, the collected mean load of WILD data can be assessed through a long-term
analytical approach, to verify the calibration procedure. For instance, the mean load distributions
of two selected cars of the MNR fleet (same car series) are compared in Figure 4-21, based on
the data derived from MNR’s WILD records. As presented in Figure 4-21, both cars have a very
symmetric mean load distribution considering both frequency and cumulative mean load values.
On the other hand, the mean load distribution of data collected by the WILD system should be
mainly bounded within the upper and lower thresholds of the wheel loads corresponding to when
the car is fully loaded or empty. It is particularly important that there should not be any mean
load less than the lower threshold.
As shown in Figure 4-21, the empty (AW0) and loaded (AW2) thresholds have been plotted on
the diagrams for car #XXXX and #YYYY to visually compare how much of the measured mean
loads are bounded within the upper and lower thresholds. The data shows that 17.5 percent (car
#XXXX) and 17.7 percent (car #YYYY) are above the upper load thresholds. The expected error
band on the WILD system measurements is 10 percent; thus, the measurements for these two
cars fall outside the expected error band. The difference of 7.5 percent and 7.7 percent,
respectively, is the actual error. Only 3.3 percent and 3.4 percent of the measured data are
distributed out of the lower load thresholds, for car #XXXX and #YYYY, respectively.

77
 Figure 4-21. Mean load distribution of two selected cars (#XXXX on left and #YYYY on
right) based on WILD data collected during 2010–2015
Overall, we can conclude that 10.8 percent (car #XXXX) and 11.1 percent (car #YYYY) of the
collected mean loads by WILD detectors were distributed beyond the acceptable levels of the
cars when they are empty and fully loaded (with considering ±10% margin error). Thus, it might
be useful to review the calibration algorithm and parameters of WILD system. As a result, such
modification on calibration procedure can provide a more reasonable mean load distribution
against the upper and lower load thresholds.

78
Appendix A.
WPD Calibration Record Requirements

Figure A-1. A draft of WPD calibration record requirements, issued by AAR [73]

79
Appendix B.
MNR’s Faults Page with WILD Alert

Figure B-1. GUI Based Interface for WILD Related Queries at MNR, Courtesy by MNR,
2015 [1]

80
Appendix C
MNR’s Mechanical Maintenance Alert MMA # 292

Figure C-1. MNR’s mechanical maintenance alert (MMA) # 292, Courtesy by MNR, 2015
[1]

81
Appendix D
Cross-sections of HBD and HWD Systems

Figure D-1. Location of HBD scanners about 7.25 inches from the gauge line [44]

Figure D-2. Location of HWD scanners to scan about 4 inches above the rail [44]

82
5. References

References
[1] Sharma & Associates, "An Overview of Wayside Systems at Metro-North Railroad – Phase
I (Draft)," FRA, Washington, DC, 2015.
[2] K. V. Kumar, "Concept and Working of WILD," India, 2015.
[3] Voestalpine, "Efficient Detection of Dragging Equipment," Voestalpine Signaling,
Germany, 2014.
[4] Ansaldo, "Weigh-In Motion (WIM) Wheel Impact Load Detector (WILD)," Ansaldo STS,
Italy, 2015.
[5] LBFoster, "Rail Condition Monitoring by Salient Systems, Weigh-In Motion," 2016.
[Online].
[6] LBFoster, "Hunting Truck Detector," LBFoster, USA, 2016.
[7] Sharma & Associates, "Effectiveness of Wayside Detector Technologies on Train
Operation Safety," FRA, Washington, DC, 2013.
[8] S. Kalay and M. Witte, "Recent Advances in Automated Wagon Health Monitoring and Its
Effects on Safety in North America," in IHHA 2015 Conference, Perth, Australia, 2015.
[9] Mermec, "MERMEC Monitoring Systems," 2013. [Online].
[10] Voestalpine, "Diagnostic and Monitoring Technologies," Voestalpine Signaling, Germany,
2015.
[11] DeltaRail, "WheelChex Commissioning Procedure," DeltaRail, UK, 2009.
[12] DeltaRail, "Data Verification and System Calibration Procedure," DeltaRail, UK, 2010.
[13] AAR, "Nationwide Wayside Detector System," 2015. [Online].
[14] TTC, "TADS, Trackside Acoustic Detection System," 2015. [Online].
[15] H. Braren, "Wayside Detection - Component Interactions and Composite Rules," Rail
Transportation Division Fall Conference (RTDF), Fort Worth, TX, 2009.
[16] TTC, "Trackside Acoustic Detection System," AAR, Pueblo, CO, 2016.
[17] AAR, "AAR Manual of Standards and Recommended Practices Sensors; Acoustic Roller
Bearing Detector Level-1 Indications, Standard S-6000," AAR, Washington, DC, 2012.
[18] T. Sultana, "Safety Impact of Wayside Detector Systems, Systems, Improved Automated
Equipment Inspection," in 22nd Annual AAR Reserach Review, Pueblo, CO, 2017.
[19] Railadvisor, "RailBAM® Bearing Acoustic Monitor," 2014. [Online]
[20] TrackIQ, "Wayside Monitoring Systems, Global product and software solutions for the rail
industry," TrackIQ, Adelaide, Australia, 2014.
[21] TrackIQ, "RailBAM® Bearing Acoustic Monitor," TrackIQ, Adelaide, Australia, 2014.
[22] Siemens, "Siemens Rail Systems Helping the UK stay on the right track to success,"
Siemens, Germany, 2013.
[23] C. Southern, A. D. Rennison and U. Kopke, "RailBAM® - An Advanced Bearing Acoustic
Monitor: Initial Operational Performance Results," in Conference On Railway Enginerring,
2004.

83
 References
[24] Wabtec, "RailBAM - Bearing Acoustic Monitor," Wabtec, 2017. [Online].
[25] M. W. Anish Poudel, "Effectiveness of Automated Inspection Systems to Detect Vertical
Split Rim and Shattered Rim Defects," in 22nd Annual AAR Reserach Review-TTC,
Pueblo, CO, 2017.
[26] Nordco, "Wayside Cracked Wheel Detection System," Nordco Inc., Connecticut, USA,
2013.
[27] FRA, "Automated Cracked Wheel Detection," 2017. [Online].
[28] M. Witte, "Automated Cracked Wheel Detection," in AAR Annual Review, Pueblo, CO,
2013.
[29] AAR, "AAR Manual of Standards and Recommended Practices Sensors; Detector
Calibration and Validation Requirements, Standard S-6101," AAA, Washington, DC,
2015.
[30] Progress-Rail, "Dragging Equipment and Derailment Detection," Progress-Rail Services,
Alabama, USA, 2016.
[31] STC, "iCube, Wayside Monitoring System, User's Guide," Southern Technologies
Corporation (STC), USA, 2014.
[32] Inspiredsystems.com.au, "Dragging Equipment Detector," Inspiredsystems, 2017. [Online].
[33] J. Cline, "Improved Hot Box Detector Performance," in 19th Annual AAR Reserach
Review, Pueblo, CO, 2014.
[34] S. Kalay, "AAR Strategic Research Implementation Safety and Efficiency," in 15th Annual
AAR Research Review, Pueblo, CO, 2010.
[35] STC, "SmartScan, Catalog of Wayside Information Systems Products and Applications
Software," Southern Technologies Corporation, USA, 2004.
[36] E. Eisenbrand, "Hot box detection in European railway networks," RTR Special, pp. 2-11,
2011.
[37] Voestalpine, "Phoenix MB Intelligent Rolling Stock Monitoring, Hot Box and Hot Wheel
Detection with Multi Beam Technology," Voestalpine Signaling, Germany, 2014.
[38] T. Sultana and S. Belport, "Evaluation of Improved Hot Bearing Detectors," in 22nd
Annual AAR Research Rview, Pueblo, CO, 2017.
[39] T. Sultana and L. Stabler, "Condition-Based Maintenance with Detectors," in IHHA, Perth,
Australia, 2015.
[40] STC, "SmartScanNG Generic User's Guide," Southern Technologies Corporation (STC),
USA, 2006.
[41] AAR, "Inoperative Brake Indications as Determined by Wayside Detectors, Standard S-
6031," Association of American Railroads (AAR), Washington, DC, 2016.
[42] WID, "Truck Condition Monitoring," Wayside Inspection Devices (WID), Canada, 2016.
[43] IHHA, Guidelines to Best Practices For Heavy Haul Railway Operations: Management of
the Wheel and Rail Interface, Virginia Beach, VA: Simmons-Boardman Books, Inc., 2015.
[44] P. Bladon, et al., "The Challenges of Integrating Novel Wayside Rolling Stock Monitoring
Technologies, A Case Study," in IHHA 2015 Conference, Perth, Australia, 2015.

84
 References
[45] G., Izbinsky, et al., "Monitoring Bogie Performance on Straight Track, Part 2. Tracking
Error and Shift," in IHHA, Sweden, 2007.
[46] Railinc, "Asset Health Data Summaries," Railinc, Cary, NC, 2015.
[47] Railinc, "Equipment Health Management System (EHMS) User Guide," Railinc, North
Carolina, 2015.
[48] H. Tournay, "Use of Wayside Detection for Rolling Stock Performance Monitoring and
Maintenance," in IHHA, Sweden, 2007.
[49] Li R. Cheng, "Vehicle Hunting & Its Effect on North American Railroad Operations," in
IHHA, Kiruna, Sweden, 2007.
[50] Vortok, "Multi-Sensor, Multi-Parameter Sensor for Monitoring Stress in Railways," Vortok
International, Devon, UK, 2017.
[51] M., Richard, et al., "System to Detect Truck Hunting on Freight Railroads," TRB,
Wshington, DC, 2006.
[52] T. W. Moynihan, et al., "Railway Safety Technologies," Research and Traffic Group,
Canada, 2007.
[53] LBFoster, "Truck Performance Detector (TPD)," LBFoster, Ohio, USA, 2005.
[54] Semih Kalay, et al., "The Safety Impact of Wagon Health Monitoring in North America,"
in WCRR, India, 2011.
[55] IEM, "Truck Performance Detector (TPD)," IEM Corp., USA, 2016.
[56] Ansaldo, "Fibre Optic Detection of Weight and Load Imbalances on Moving Trains,"
Ansaldo STS, Italy, 2015.
[57] Li R., Cheng, et al., "Effects of Improper Loading in Heavy Haul Operations," in IHHA,
Sweden, 2007.
[58] SchenckProcess, "Schenck Process Rail weighing systems," 2012. [Online]
[59] LBFoster, "Rail Monitoring by Salient Systems, WILD," 2016. [Online]
[60] International Engineering, "WILD-Wheel Impact Load Detector & WILD +," International
Engineering, August 2016. [Online].
[61] SchenckProcess, "MULTIRAIL WheelScan," 2013. [Online].
[62] Track IQ, "Wheel Condition Monitor (WCM)," Track IQ, Australia, 2016.
[63] Voestalpine, "ATLAS FO Precise Diagnosis of Wheel Defects and Vehicl Weights,"
Voestalpine Signaling, Germany, 2014.
[64] DeltaRail, "Asset Management," UK, 2016.
[65] J. Kesich and A. Golby, "MNR Wheel Impact Load Detection – Improved Performance at
Reduced Cost," in APTA conference, Boston, 2011.
[66] AAR, "Wheel Impact Load Detector Tests and Development of Wheel-Flat Specification,
Report no. R-829," AAR, Chicago, IL, 1993.
[67] LYNXRAIL, "Automated High-Speed Vehicle Inspection Systems," LYNXRAIL,
Australia, 2011.
[68] AEA, "Appendix 1, View, Automatic Vehicle Inspection Systems, A Product Profile by
AEA Technology Rail," in IHHA, Kiruna, Sweden, 2007.

85
 References
[69] M. Asplund and et al., "Automatic Laser Scanning of Wheel Profiles: Condition
Monitoring to Achieve Greater Capacity for Existing Infrastructure in an Extreme
Climate," in IHHA, New Delhi, India, 2013.
[70] AAR, "AAR Manual of Standards and Recommended Practices, Wheel Profile Detector
Standard S-6103," Association of Americn Railroads (AAR), Washington, DC, 2017.
[71] FRA, "Using Wheel Temperature Detector Technology to Monitor Railcar Brake System
Effectiveness," Technical Report No. DOT/FRA/ORD-13/50, Federal Railraod
Administration, Washington, DC, 2013.
[72] B. J. V. Dyk, M. S. Dersch, J. R. Edwards, J. Conrad, J. Ruppert and C. P. L. Barkan,
"Evaluation of Dynamic and Impact Wheel Load Factors and their Application for
Design," in TRB 93rd Annual Meeting, Washington, DC, 2014.
[73] B. Stratman, Y. Li and S. Mahadevan, "Structural Health Monitoring of Railroad Wheels
Using Wheel Impact," Vanderbilt University, Nashville, TN, 2006.
[74] P. Harold, D. Harrison and J. M. Tuten. USA Patent 4,701,866, 1987.
[75] A. A. S. Elsaleiby, "Wheel Imbalance Effect on the Output of Wheel Impact Load Detector
System (WILD), A Thesis," Colorado State University, Pueblo, CO, 2014.
[76] M. L. Lee and W. K. Chiu, "A comparative study on impact force prediction on a railway
track-like structure," Structural Health Monitoring, vol. 4, no. 4, pp. 355-376, 2005.
[77] S. Lechowicz and C. Hunt, "Monitoring and Managing Wheel Condition and Loading,"
Teknis, Australia, 2000.
[78] TSB, "Railway Investigation Report, R13T0060, Main-track train derailment,"
Transportation Safety Board of Canada, Ontario, 2015.
[79] FRA, "Federal Railroad Administration, Office of Safety Analysis," 2011. [Online].
[80] Nagory FOSTER, "Wheel Impact Load Detector," Nagory FOSTER, India, 2015.
[81] A. Nurmikolu, P. Salmenpera, S. Mäkitupa and K. Lane, "Statistical Analysis of Wheel
Impact Load Data and Review for Finnish Impact Load Limits," in 10th IHHA Conference,
India, 2013.
[82] R. Wiley, "Railway Fleet Management: Using Acoustic Bearing Detector (ABD) Data,"
2016.
[83] ApnaTech, " Railway Products and Solutions," Apna Technologies & Solutions, India,
2017.
[84] TTC, "AAR Strategic Research Implementation Safety and Efficiency," AAR,
Washington, DC, 2010.
[85] R. Donnelly and et al., "Application of an Acoustic Bearing Monitor in a Heavy-Haul
Environment," in IHHA, Sweden, 2007.

86
Acronyms and Abbreviations
Abbreviations & Definitions
Acronyms
ABD Acoustic Bearing Detectors
X A given parameter/variable
AW0 All Wheels Unloaded (empty transit car)
AW2 All Wheels Loaded (loaded transit car)
AMS Asset Management System
AAR Association of American Railroads
TPDG A TPD Measurement of Horizontal Force
TPDL A TPD Measurement of Lateral over Vertical Force
ACWD Automatic Cracked Wheel Detector
AEI Automatic Equipment Identification
BNSF Burlington Northern Santa Fe Railway
CDMI Calendar Day Mechanical Inspection
CN Canadian National Railway
CP Canadian Pacific Railway
CTC Centralized Train Controller/Center
CCI Certified Car Inspection
C&S Communication and Signal
DED Dragging Equipment Detector
EHMS Equipment Health Management System
FDP Fault Diagnostic Procedure
FRA Federal Railroad Administration
FMO Fleet Management Office
FactIS Fully Automated Car Train Inspection System
GCT Grand Central Terminal
GUI Graphic User Interface
g Gravity Force
GPRS General Packet Radio Service
GPS Global Position System
HBD Hot Box Detector
HWD Hot Wheel Detector
HI Hunting Index
IF Impact Factor
IMS Inspection Management System
InteRRIS® Integrated Railway Remote Information Service
K Kelvin Degree
kips Kilo-Pounds
kN Kilo-Newton
L/V Lateral Force/Vertical Force
LAHRLV Lead Axle High Rail Lateral/Vertical
LIRR Long Island Rail Road
MofE Maintenance of Equipment
MMA Mechanical Maintenance Alert

87
Abbreviations & Definitions
Acronyms
MHz Megahertz
MNCW Metro-North Commuter Railroad
MNR Metro-North Railroad
MTA Metropolitan Transportation Authority
MP Milepost
mm Millimeter
NR National Rail (Australia)
NYA New York & Atlantic Railway
NB Northbound
NEC Northeast Corridor
NA Not Available, or Not Applicable
OCC Operations Control Center
OPS Operations Service
OEM Original Equipment Manufacturer
QMP Qualified Maintenance Personnel
lbs. Pounds
RFID Radio-Frequency Identification
RailBAM® Railway Bearing Acoustic Monitor
RTC Rail Traffic Controller
ROC Receiver Operating Characteristics
RD&T Research, Development & Technology
SMS Short Message System
SB Southbound
STD Standard Deviation
TADS® Trackside Acoustic Detection System
TTC Transportation Technology Center
TBOGI Truck Bogie Optical Geometry Inspection
THD Truck Hunting Detector
TPD Truck Performance Detector
UMLER® Universal Machine Language Equipment Register
UP Union Pacific Railroad
DOT U.S. Department of Transportation
VSR Vertical Split Rim
WIM Weigh-In Motion Detector
WILD Wheel Impact Load Detector
WPMS Wheel Profile Measurement System
WTD Wheel Temperature Detector
WIFI Wireless Fidelity

88