Condition monitoring of rail infrastructure is an important task to ensure the safety and ride quality. The increasing travel demands of the rail network due to higher miles traveled requires regular monitoring of the infrastructure and efficient processing of the data for timely decision-making. Despite the regular data collection on different parameters such as acceleration and track geometry, the data processing is commonly performed to document the track performance and maintenance without further knowledge discovery to realize all the potential from historical data. Motivated by the wealth of historical track data in practice, this paper investigates the feasibility of using onboard data that is repeatedly collected over a period of time on a segment of track to potentially identify changes to the track. The proposed approach has been envisioned to learn from repeated historical time-series data to identify both the location and timing of unexpected changes to the track system. To account for stochastic nature of the collected data, associated with the temporal mismatch between the time-series across different inspection runs, we propose a framework by adopting the concept of Matrix Profile without relying on time series synchronization. The approach divides the entire data into smaller track segments, performs extensive similarity search of time-series signatures, and associate locations with higher dissimilarity to changes of the track either due to maintenance or a potential defect. To demonstrate the efficacy and potential of the method, evaluation on both synthetic data and the field geometry data from a revenue-service Class I railroad has been conducted. The findings provide promising results in predicting the location of track changes with a reasonably high degree of certainty, with an automated offline analysis.

The railroad industry currently utilizes two wayside detection systems to monitor the health of freight railcar bearings in service: The Trackside Acoustic Detection System (TADS™) and the wayside Hot-Box Detector (HBD). TADS™ uses wayside microphones to detect and alert the conductor of high-risk defects. Many defective bearings may never be detected by TADS™ since a high-risk defect is a spall which spans more than 90% of a bearing’s raceway, and there are less than 20 systems in operation throughout the United States and Canada. Much like the TADS™, the HBD is a device that sits on the side of the rail-tracks and uses a non-contact infrared sensor to determine the temperature of the train bearings as they roll over the detector. These wayside detectors are reactive in the detection of a defective bearing and require emergency stops in order to replace the wheelset containing the defective bearing. These costly and inefficient train stoppages can be prevented if a proper maintenance schedule can be developed at the onset of a defect initiating within the bearing. This proactive approach would allow for railcars with defective bearings to remain in service operation safely until reaching scheduled maintenance. Driven by the need for a proactive bearing condition monitoring system in the rail industry, the University Transportation Center for Railway Safety (UTCRS) research group at the University of Texas Rio Grande Valley (UTRGV) has been developing an advanced onboard condition monitoring system that can accurately and reliably detect the onset of bearing failure using temperature and vibration signatures of a bearing. This system has been validated through rigorous laboratory testing at UTRGV and field testing at the Transportation Technology Center, Inc. (TTCI) in Pueblo, CO. The work presented here builds on previously published work that demonstrates the use of the advanced onboard condition monitoring system to identify defective bearings as well as the correlations developed for spall growth rates of defective bearing outer rings (cups). Hence, the system uses the root-mean-square (RMS) value of the bearing’s acceleration to assess its health. Once the bearing is determined to have a defective outer ring, the RMS value is then used to estimate the defect size. This estimated size is then used to predict the remaining service life of the bearing. The methodology proposed in this paper can prove to be a useful tool in the development of a proactive and cost-efficient maintenance cycle for railcar owners.

The railroad industry currently utilizes two wayside detection systems to monitor the health of freight railcar bearings in service: The Trackside Acoustic Detection System (TADS™) and the wayside Hot-Box Detector (HBD). TADS™ uses wayside microphones to detect and alert the conductor of high risk defects. Many defective bearings may never be detected by TADS™ due to the fact that a high risk defect is considered a spall which spans more than 90% of a bearing’s raceway, and there are less than 20 systems in operation throughout the United States and Canada. Much like the TADS™, the HBD is a device that sits on the side of the rail tracks and uses a non-contact infrared sensor to determine the temperature of the train bearings as they roll over the detector. The accuracy and reliability of the temperature readings from this wayside detection system have been concluded to be inconsistent when comparing several laboratory and field studies. The measured temperatures can be significantly different from the actual operating temperature of the bearings due to several factors such as the class of railroad bearing and its position on the axle relative to the position of the wayside detector. Over the last two decades, a number of severely defective bearings were not identified by several wayside detectors, some of which led to costly catastrophic derailments. In response, certain railroads have attempted to optimize the use of the temperature data acquired by the HBDs. However, this latter action has led to a significant increase in the number of non-verified bearings removed from service. In fact, about 40% of the bearings removed from service in the period from 2001 to 2007 were found to have no discernible defects. The removal of non-verified (defect-free) bearings has resulted in costly delays and inefficiencies. Driven by the need for more dependable and efficient condition monitoring systems, the University Transportation Center for Railway Safety (UTCRS) research team at the University of Texas Rio Grande Valley (UTRGV) has been developing an advanced onboard condition monitoring system that can accurately and reliably detect the onset of bearing failure. The developed system currently utilizes temperature and vibration signatures to monitor the true condition of a bearing. This system has been validated through rigorous laboratory testing at UTRGV and field testing at the Transportation Technology Center, Inc. (TTCI) in Pueblo, CO. The work presented here provides concrete evidence that the use of vibration signatures of a bearing is a more effective method to assess the bearing condition than monitoring temperature alone. The prototype bearing condition monitoring system is capable of identifying a defective bearing with a defect size of less than 6.45 cm 2 (1 in 2 ) using the vibration signature, whereas, the temperature profile of that same bearing will indicate a healthy bearing that is operating normally.

Prevention of railroad bearing failures, which may lead to catastrophic derailments, is a central safety concern. Early detection of railway component defects, specifically bearing spalls, will improve overall system reliability by allowing proactive maintenance cycles rather than costly reactive replacement of failing components. A bearing health monitoring system will provide timely detection of flaws. However, absent a well verified model for defect propagation, detection can only be used to trigger an immediate component replacement. The development of such a model requires that the spall growth process be mapped out by accumulating associated signals generated by various size spalls. The addition of this information to an integrated health monitoring system will minimize operation disruption and maintain maximum accident prevention standards enabling timely and economical replacements of failing components. An earlier study done by the authors focused on bearing outer ring (cup) raceway defects. The developed model predicts that any cup raceway surface defect (i.e. spall) once reaching a critical size (spall area) will grow according to a linear correlation with mileage. The work presented here investigates spall growth within the inner rings (cones) of railroad bearings as a function of mileage. The data for this study were acquired from defective bearings that were run under various load and speed conditions utilizing specialized railroad bearing dynamic test rigs owned by the University Transportation Center for Railway Safety (UTCRS) at the University of Texas Rio Grande Valley (UTRGV). The experimental process is based on a testing cycle that allows continuous growth of railroad bearing defects until one of two conditions are met; either the defect is allowed to grow to a size that does not jeopardize the safe operation of the test rig, or the change in area of the spall is less than 10% of its previous size prior to the start of testing. The initial spall size is randomly distributed as it depends on the originating defect depth, size, and location on the rolling raceway. Periodic removal and disassembly of the railroad bearings was carried out for inspection and defect size measurement along with detailed documentation. Spalls were measured using optical techniques coupled with digital image analysis, as well as, with a manual coordinate measuring instrument with the resulting field of points manipulated in MatLab™. Castings were made of spalls using low-melting, zero-shrinkage bismuth-based alloys, so that a permanent record of the spall geometry and its growth history can be retained. The main result of this study is a preliminary model for spall growth, which can be coupled with bearing condition monitoring tools that will allow economical and effective scheduling of proactive maintenance cycles that aim to mitigate derailments, and reduce unnecessary train stoppages and associated costly delays on busy railways.

Prevention of bearing failures which may lead to catastrophic derailment is a major safety concern for the railroad industry. Advances in bearing condition monitoring hold the promise of early detection of bearing defects, which will improve system reliability by permitting early replacement of failing components. However, to minimize disruption to operations while providing the maximum level of accident prevention that early detection affords, it will be necessary to understand the defect growth process and try to quantify the growth speed to permit economical, non-disruptive replacement of failing components rather than relying on immediate removal upon detection. The study presented here investigates the correlation between the rate of surface defect (i.e. spall) growth per mile of full-load operation and the size of the defects. The data used for this study was acquired from defective bearings that were run under various load and speed conditions utilizing specialized railroad bearing dynamic test rigs operated by the University Transportation Center for Railway Safety (UTCRS) at the University of Texas Rio Grande Valley (UTRGV). Periodic removal and disassembly of the railroad bearings was carried out for inspection and defect size measurement and documentation. Castings were made of spalls using low-melting, zero shrinkage Bismuth-based alloys so that a permanent record of the full spall geometry could be retained. Spalls were measured using optical techniques coupled with digital image analysis and also with a manual coordinate measuring instrument with the resulting field of points manipulated in MatLab™ and Solidworks™. The spall growth rate in area per mile of full-load operation was determined and, when plotted versus spall area, clear trends emerge. Initial spall size is randomly distributed as it depends on originating defect depth, size, and location on the rolling raceway. The growth of surface spalls is characterized by two growth regimes with an initial slower growth rate which then accelerates when spalls reach a critical size. Scatter is significant but upper and lower bounds for spall growth rates are proposed and the critical dimension for transition to rapid spall growth is estimated. The main result of this study is a preliminary model for spall growth which can be coupled to bearing condition monitoring tools to permit economical scheduling of bearing replacement after the initial detection of spalls.

In the railroad industry, monitoring the condition of key components such as bearings and wheels is vital to ensure the safe transport of goods and commodities. Bearing seizures are amongst the most dangerous types of failures experienced by trains because they occur unexpectedly and may lead to costly derailments. Current bearing health monitoring techniques include tracking the temperature and acoustic emissions given by the bearings. Although temperature histories of railroad tapered roller bearings are readily available, the literature does not provide information relating the temperature profiles to the severity of the bearing defect. The study presented here investigates the correlation between temperature profiles and bearing defect severity measured by the size of spalls present on bearing outer (cup) and inner (cone) rings. The temperature data used for this study was acquired from defective and healthy bearings that were run at various operating load and speed conditions. The data presented here provides the railroad industry with a greater understanding of the thermal behavior of defective bearings, which can be used to assess the future needs of bearing condition monitoring systems.

Draft gears and cushioning devices absorb energy during relative motion between two coupled railcars. This may occur in the form of buff and draft forces during over-the-track operation or impact forces during a coupling event in a switching yard. During such events, the end-of-car device absorbs energy that would otherwise be transmitted and possibly cause damage to the car body and/or lading. Health and performance of the end-of-car device is therefore of concern to shippers and car owners. Damaged cushioning units may be identified by visual inspection if the device is leaking hydraulic fluid or by a unit condition indicator, but may not be noticed until damage has been caused to the car and/or lading. This paper presents the results of a study on the detection of a faulty end-of-car cushioning unit. A boxcar with 15 inch travel cushioning units installed on both ends was deployed for field service. The draft systems were inspected prior to the car being deployed, and although both ends of the car initially showed some signs of wear, there was evidence of heavier wear as well as signs of leaking on the B-end cushioning unit. The goal of the field trial was to determine whether or not the B-end of the car would produce higher magnitude acceleration events than the A-end during revenue service. Three wireless onboard monitoring systems, each with GPS and an accelerometer, were used to determine car location and measure car body accelerations. High car body accelerations were recorded and studied for a three month period of time. The results show that the onboard monitoring systems successfully detected more frequent and higher magnitude acceleration events on the B-end of the car.

An advanced technology for measuring track substructure conditions has been developed by Sol Solution in France. This technology, referred to as the PANDA® and Geo-Endoscopy® technique, comprises driving a variable energy cone penetration device into the track substructure using an instrumented anvil to collect strength (and modulus by correlation) profile with depth. Condition monitoring of the track substructure layers is accomplished through insertion of a camera into the same hole, also called Geoendoscopy®. The tests are commonly faster than Dynamic Cone Penetration (DCP) testing and cause negligible damage to the track substructure with the use of light and portable devices. This paper presents data from PANDA® and Geo-Endoscopy® testing efforts recently carried out under the scope of an ongoing research study at the University of Illinois. The primary objective was to relate the substructure layer properties to plastic layer deformations measured through advanced geotechnical instrumentation. Combined analysis of the PANDA®, Geo-Endoscopy®, and field instrumentation data has been used to highlight the reliability of this innovative technique towards improved evaluation of track substructure layer conditions.

North American Railroads and the United States Department of Transportation (US DOT) Federal Railroad Administration (FRA) require periodic inspection of railway infrastructure to ensure the safety of railway operation. Tracks that are subjected to heavy haul or high-speed traffic necessitate frequent inspection and more stringent maintenance requirements, but present railroads with less time to accomplish it. The international railroad community has undertaken significant research to develop innovative applications for advanced technologies with the objective of improving the process of visual track inspection. Some of these technologies are currently in use or under development for a variety of railroad inspection tasks, both wayside and mobile. This paper presents an overview of different systems for condition monitoring applications for inspection of railway components. These technologies, in conjunction with defect analysis and comparison with historical data, will enhance the ability for longer-term predictive assessment of the health of the track system and its components, more informed and proactive maintenance strategies, and improved understanding of track structure degradation and failure modes.

Automated wayside inspection systems have been developed to monitor rolling stock condition and facilitate predictive maintenance strategies. Improving the effectiveness of railcar condition monitoring has the potential for significant cost savings as a result of reduced derailments and in-service failures. This paper presents an economic analysis of the costs associated with train accidents and mainline delay due to defective railcar components. We consider track and equipment damages as well as train delay costs. We use dispatch simulation software to determine the impact of in-service failures on mainline capacity at varying traffic levels. This analysis provides a basis for the comparison of current railcar inspection practices with new integrated practices incorporating automated inspection systems.

If the contact between the wheel and rail occurs at the flange, a high degree of slip is inherent. This can result in excessive wear and the potential for rolling contact fatigue cracking. Multi-body dynamics software is useful for predicting the bogie and track characteristics that may lead to flange contact occuring. However, there is a shortage of experimental tools available to validate such models or to assist in vehicle and track condition monitoring. In this study, the feasibility of a new approach is investigated. The method is based on an ultrasonic sensor mounted on the wheel. The concept is that the sensor emits an ultrasonic pulse which is designed to impinge on the wheel flange. If there is no contact the pulse is fully reflected back at the flange and picked up by the same sensor. If flange contact takes place, a proportion of the pulse amplitude will be transmitted into the rail. The signal reflected back to the sensor is therefore reduced. The amount by which this signal reduces indicates how much flange contact occurs. Test specimens were cut from sections of wheel and rail, and a 2MHz ultrasonic contact transducer was bonded onto the wheel in a position best suited to detect the flange contact. The specimens were pressed to together in a bi-axial loading frame to generate differing degrees of head and flange contact. The reflected signal was monitored as the normal and lateral loads were varied. It proved possible not only to detect the onset of flanging, but also to record a signal that varied monotonically with both normal and lateral applied load. A map of reflected ultrasound against the applied loading is presented.

In 2006, Amtrak instituted a program that coupled Reliability Centered Maintenance (RCM) with Root Cause Analysis (RCA) to increase rolling stock reliability. The RCM-RCA program at Amtrak combined features of RCM and RCA into a unique methodology to achieve success. A case study of the High Horsepower electric locomotive (HHP-8) power module failure RCA project is described in this paper. The initial focus of the RCA was a persistently high failure rate of power modules. Power modules are expensive to purchase and labor intensive to replace. Additionally, power module failures take locomotives out of revenue service. Partially as a result of the HHP-8 Power Module RCA project, and the implemented recommendations, the availability of the HHP-8 locomotive fleet has improved by 80%. Changes were made to the maintenance, condition monitoring, and troubleshooting procedures for HHP-8 as a result of the RCA project. The physical, human, and latent root causes for the high failure rate of HHP-8 Power Modules, were determined and addressed during this six month effort. A cross-functional team of AMTRAK personnel worked closely with reliability consultants and the original equipment manufacturers to determine the various root causes. This case study is a valuable example of how the hands-on nature of a rigorous RCA coupled with the methodology of RCM can result in significant improvements to reliability and availability.

This paper presents the research work carried out at Oxford University on condition monitoring of railway point machines. The developed condition monitoring system includes a variety of sensors for acquiring trackside data related to different parameters. Key events to be logged include time stamping of points operation, opening and closing of case cover associated with a points machine, insertion and removal of a hand-crank, loss of supply current and the passage of a train. The system also has built-in Web functions. This allows a remote operator using Internet Explorer to observe the condition of the point machine at any time, while the acquired data can be downloaded automatically for offline analysis, providing more detailed information on the health condition of the monitored point machine. A short daily condition report message can also be sent to relevant staff via email. At last the experience are reported on the four trackside installed systems.

Vibration signatures of defective roller bearings on railroad freight cars were analyzed in an effort to develop an algorithm for detecting bearing defects. The effort is part of a project to develop an on-board condition monitoring system for freight trains. The Office of Research and Development of the Federal Railroad Administration (FRA) is sponsoring the project. The measurements were made at the Transportation Technology Center (TTC) in Pueblo, CO on July 26 – 29, 1999 during the Phase III Field Test of the Improved Wayside Freight Car Roller Bearing Inspection Research Program sponsored by FRA and the Association of American Railroads (AAR). Wheel sets with specific roller bearing defects were installed on a test train consisting of 8 freight cars designed to simulate revenue service. The consist also contained non-defective roller bearings. Accelerometers were installed on the inboard side of the bearing adapters to measure the vibration signatures during the test. Signatures of both defective and non-defective bearings were recorded. The data were recorded on Sony Digital Audio Tape (DAT) Recorders sampling at a rate of 48 K samples per second. We used both ordinary and envelope spectral analysis to analyze the data in an effort to detect features that could be related to known defects. The spectra of non-defective bearings show no remarkable features at bearing defect frequencies. In general, the ordinary spectra of defective bearings do not exhibit remarkable features at the bearing defect frequencies. In contrast, the envelope spectra of defective bearings contain a number of highly resolved spectral lines at these frequencies. In several cases the spectral lines could be related to specific bearing defects. Based on the analysis performed to date, the envelope spectrum technique provides a promising method for detecting defects in freight car roller bearings using an on-board condition monitoring system.

The Office of Research and Development of the Federal Railroad Administration (FRA) is sponsoring a revenue service demonstration of an on-board condition monitoring system for freight trains. The objective of the system is to improve railroad safety and efficiency through continuous monitoring of mechanical components to detect defects before they cause breakdowns and accidents. The project, which commenced in June 1999, is part of the Rolling Stock Program Element in FRA’s Five-Year Strategic Plan for Railroad Research, Development and Demonstrations [1]. Science Applications International Corporation (SAIC) and Wilcoxon Research (WR) developed the prototype system in 2000. It was installed on a test vehicle provided by the Research and Tests Department at Norfolk Southern Corporation and was tested during the period Nov. 2000–Nov. 2001. In fall 2003 the monitoring system was installed on five hopper cars provided by Southern Company Services and is currently being tested in revenue service operation on a coal train operating on a Norfolk Southern route in Alabama between a coalmine northwest of Birmingham and Gaston Steam Plant in Wilsonville, AL. In this paper we discuss the performance of the monitoring system during the revenue service demonstration. The potential benefits of the technology are also addressed.