2021

Tianci Gao, Southwest Jiaotong University
Zihan Li, Southwest Jiaotong University
Qihang Wang, Southwest Jiaotong University
Kanghua Yang, Southwest Jiaotong University
Cuiping Yang, Southwest Jiaotong University
Chenzhong Li, Southwest Jiaotong University
Ping Wang, Southwest Jiaotong University
Qing He, Southwest Jiaotong University

A small-radius curve is one of the weaker areas of a railway line, bearing a large amount of lateral rail-wheel interactions that often cause the rail to wear out relatively quickly past wear threshold levels. Therefore, providing a reasonable estimate for the renewal period due to rail wear in small-radius curves can not only improve the safety of railway operations but also substantially reduce maintenance cost. In light of this, this paper presents the development of a data and mechanics integrated approach (DMIA) to analyze three different types of rail wear measurements recorded from 2017 to 2018. First, the Archard wear model (a mechanics-based model) is adopted to calculate the wear growth chart under different radii. Second, based on the growth chart, the rail lifetime is estimated in million gross tonnage (MGT) of over-threshold rail wear readings, immediately as the rails reach the wear threshold. Third, considering the influence of different curve radii and the inner or outer rails on rail wear, an accelerated failure time (AFT) model (a data-driven model) is used to analyze the heterogeneity of the corrected measurement data. A survival probability is obtained with increasing gross tonnage for different curve radii. Finally, the expected MGT is calculated for rail wear-based renewal for curves with different radii, and the obtained values are compared with the corresponding values in the existing standard.

Ryan Harrington, University of Illinois
Richard Fox-Ivey, Railmetrics
Arthur Lima, University of Illinois, Urbana-Champaign
Thanh Nguyen, Railmetrics
Van Nguyen, Railmetrics
Marcus Dersch, University of Illinois, Urbana-Champaign
John Laurent, Railmetrics
John Edwards, University of Illinois, Urbana-Champaign

FRA-mandated railroad track inspections result in significant labor costs and occupy valuable network capacity.  These factors, combined with advancements in the field of machine vision, have encouraged transition from human visual inspections to machine-based alternatives.  While commercial machine vision technologies for railway inspection currently exist, many depend on human interpretation of captured information which suffers similar limitations to fully manual inspections.  Automated analysis approaches which deliver objective results are also available in the industry.  However, they are limited to a “pass/fail” approach through the detection of components which fail to meet maintenance or safety thresholds, as opposed to being able to detect subtle changes in track conditions to identify evolving problems.  To overcome the limitations of human interpretation and simple “pass/fail” defect-finding, this paper describes field deployment and validation of a system that pairs three-dimensional (3D) machine vision with automated change detection technology.  The change detection approach uses a combination of traditional image processing and a deep convolutional neural network (DCNN) to accurately characterize network conditions between repeat runs before analyzing differences.  This paper first provides context for the current automated track inspection technology and discusses the applicability of change detection.  Then it discusses the process for 3D image capture and how DCNNs were trained with these 3D images.  Finally, it compares the trained DCNNs to an expert human inspector.  Results from this study suggest that this technology can successfully identify present, broken, and missing spikes and fasteners with percent accuracies in excess of 98%.

Diptojit Datta, University of California, San Diego
Ranting Cui, University of California, San Diego
Francesco Lanza di Scalea, University of California, San Diego
Robert Wilson, Federal Railroad Administration (FRA)

Internal defects in rails are a major cause of derailments and train accidents around the world. Existing techniques for inspecting internal flaws in rails operate at speeds of up to 30 mph which is considerably slower than revenue speeds (~ 60 mph). Lower inspection speeds result in disruptions to normal traffic which is undesirable. This study presents a high-speed rail inspection technique that has the potential of detecting internal rail flaws at regular revenue speeds and could complement traditional rail inspections. The technology uses a non-contact, passive ultrasonic sensing technique that utilizes air-coupled transducers that pick up the ultrasonic guided waves generated by the wheels of the locomotive into the rail and hence does not require a controlled source of excitation. The acoustic transfer function of the rail between two points is extracted through a normalized cross-correlation operator with additional processing to remove uncorrelated noise. The features from this transfer function are statistically analyzed to determine if a rail segment has existing damage. A prototype with multiple pairs of air-coupled capacitive sensors and a data-acquisition system programmed in LabView Real-time was developed to perform non-contact, high-speed rail inspection in real-time. From field tests conducted at the Transportation Technology Center (TTC) in Pueblo, CO, the performance of the system was evaluated using Receiver Operating Characteristic (ROC) curves for a range of different operational parameters such as speed, signal-to-noise ratio, baseline distribution and redundancy from multiple runs.

Bin Feng, University of Illinois, Urbana-Champaign
Yuamar Basarah, University of Illinois, Urbana-Champaign
Qiusheng Gu, Zhejiang University
Xiang Duan, Zhejiang University
Xuecheng Bian, Zhejiang University
Erol Tutumluer (tutumlue@illinois.edu), University of Illinois, Urbana-Champaign
Youssef Hashash, University of Illinois, Urbana-Champaign
Hai Huang, Pennsylvania State University

Understanding ballast layer dynamic response and long-term behavior under moving train loads is essential for optimization of railway track performance. Advanced testing under controlled laboratory environment and measurement of track responses via multitude of installed sensors are needed to develop validated computer models and simulation tools suitable for analyzing full-scale ballasted track under dynamic loading. Constructed in a large rectangular metal frame, the laboratory model consisted of full-scale track with eight crossties, ballast, subballast, and embankment. Three different speeds and axle load configurations were applied sequentially onto the full-scale ballasted track using eight actuators, which realistically captured both slow moving heavy freight and high-speed passenger train loads. Vibration velocities were measured at different locations on ballast surfaces and crossties. The effect of train speed increase on vibration velocity was significant, especially for tie responses; but the effect of train axle load increase on ballast and tie vibration velocity was minimal. Dynamic soil stress increase was noticeable as the axle load increased but not as substantial as in the case of increased train speed. At 300 km/h speed, the highest permanent deformation accumulation rates were realized. “SmartRock” advanced ballast particle sensors captured very high particle accelerations for the speed of 300 km/h and yet did not indicate much difference in accelerations when axle loads were increased at a slower 100 km/h. Finally, a realistic DEM model of the ballast layer successfully reproduced tie and ballast vibration velocities in terms of both trends and magnitudes when compared to laboratory measured results.

Shuai Yu (sqy5325@psu.edu), Pennsylvania State University, University Park
Shihui Shen, Pennsylvania State University, Altoona
Hai Huang, Pennsylvania State University
Cheng Zhang, Pennsylvania State University, University Park

Considerable variation in the vertical stiffness or displacement at the bridge approach can cause railway track transition problem. This vertical displacement gaps would result in amplification of the dynamic force and frequency, and gradually degrade the serviceability of the railway track. Many strategies, either focusing on modifying the track component or making changes to the entire structure, were used to mitigate the transition problems. In particular, asphalt trackbed has been proposed as a structural mitigation method to provide additional support to the ballast and protect subgrade. However, its effect of reducing dynamic impact at the bridge approach is limited because asphalt mixture has limited range of stiffness and modulus and cannot make enough alleviations to the entire track structure. Therefore, this paper aims to develop an engineered semi-flexible composite mixture (SFCM) design to replace the asphalt mixture and develop an implementation method for the SFCM material to mitigate the transition problem. It was found from the experiment that the SFCM is a viscoelastic material with wider modulus range, and its stiffness and modulus could adjust with the air voids of the core structure and entire mixture. Track analysis using a 2.5D sandwich model was conducted to simulate the effects of the structure and material on the responses of the railway track under the dynamic loads. Based on the findings from the track analysis, a four-segment transition zone design was proposed, which would allow railway track to experience a smooth transition at the bridge approach.

Bin Feng, University of Illinois, Urbana-Champaign
Zhongyi Liu, University of Illinois
Erol Tutumluer (tutumlue@illinois.edu), University of Illinois, Urbana-Champaign
Hai Huang, Pennsylvania State University

Ballasted track substructure is designed and constructed to provide uniform crosstie support and aid in the distribution of superstructure wheel loads. Poor and nonuniform support conditions can cause excessive crosstie vibration which will negatively impact the crosstie flexural bending behavior. Furthermore, ballast-tie gaps and large contact forces formed in the particulate nature ballast will result in faster ballast layer degradation and settlement accumulation. Inspection of crosstie support condition is therefore necessary while very challenging to implement using current existing methods and technologies. This paper introduces an innovative data-driven crosstie support condition prediction system named deep residual neural network and its laboratory application on a newly constructed full-scale ballasted track. Discrete Element Method (DEM) is leveraged to provide training and testing datasets for the proposed prediction model. K-means clustering method is applied on ballast layer to determine most informative representative particles and show additional insights on layer zoning for dynamic behavior trends over traditional ballast layer subsections. When provided with DEM simulated particle vertical accelerations, the developed deep residual neural network could achieve 100% training and 95.8% testing accuracy. Feeding the vertical acceleration measurements captured by advanced “SmartRock” sensors from the full-scale ballasted track laboratory experiment, the trained model could successfully reach a high accuracy of 92.0%. Based on the developed deep learning approach and the research findings presented in this paper, the innovative crosstie support condition prediction system is envisioned to provide railroaders accurate, timely, and repeatable inspection and monitoring opportunities without disrupting operation of railway network.