An Intelligent Transportation Systems Approach to Railroad Infrastructure Performance Evaluation

Railroads spend billions of dollars on infrastructure maintenance and condition monitoring each year [1]. Federal laws currently specify the type and regularity of full track inspections. Railroad companies deploy relatively slow and expensive methods using human inspectors and automated inspection vehicles to search for possible defects. The expense and labor requirement of these existing non-destructive evaluation (NDE) methods limit their ability to scale for continuous and networkwide monitoring for risk mitigation and safety improvements. Hence, the industry could save billions of dollars if sensors aboard regular rolling stocks could screen the infrastructure for faults automatically and continuously. Such a solution would provide asset managers with an ability to focus inspections on areas of high fault likelihood and severity without closing lines to search for developing faults. Federal track safety standards require railroads to inspect all tracks in operations as often as twice weekly. However, with resources thinly stretched and the rate of defect formation escalating with traffic load-density, railroads are seeking to enhance the efficiency of inspections and maintenance of way. The current inspection practice not only decrease rail productivity by taking away track time to perform inspection and maintenance but also increase safety risk for railway inspection workers. There were 321 railroad worker fatalities from 2000 to 2015, and 132 were employees or contractors involved in track inspection and maintenance. Fault screening sensors on rolling stock will automatically and continuously provide track condition profile and characterize potential defects. Such a solution would free up more track time and capacity previously reserved for manual inspections while improving safety for railroad workers on duty. The inertial responses of a railcar are symptoms of possible track or equipment defects. Although not all defects do, a significant majority produce accelerated car movements in all directions [2]. Inertial sensors that monitor vehicle-track interactions (VTI) exist. However, they do not classify fault types or their level of severity [3]. Existing sensors generally produce alerts directly when acceleration magnitudes exceed fixed thresholds. Inspectors must then travel to the estimated locations of every alert event to search for associated faults. Technicians must also pre-configure such sensors with thresholds based on their experiences or intuition. Without an objective and consistent approach to setting thresholds, some significant defects could go unnoticed whereas minor ones could result in unnecessary and expensive inspections. The inertial responses of vehicles will vary with train speed, gross weight, suspension system design, and weather conditions. Therefore, thresholds must adapt to the specific circumstances to improve accuracy and reduce false positives. However, threshold adaptation can be complex and expensive. Adaption that uses complex algorithms on remote servers will require that the sensors support two-way communications, thus increasing their cost. The method developed in this research will not rely on adapting sensor configurations and will require only a data upload capability. The new sensors will compress and upload their geo-tagged inertial data periodically to a centralized processor. Remote algorithms will combine and process the data from multiple train traversals to identify fault symptoms, rank their severity, classify fault types, and localize their position. Fault classification will enable asset managers to allocate the appropriate specialists to scrutinize the fault location.

Language

  • English

Project

  • Status: Active
  • Funding: $226000
  • Contract Numbers:

    DTRT13-G-UTC38

  • Sponsor Organizations:

    Research and Innovative Technology Administration

    University Transportation Centers Program
    1200 New Jersey Avenue
    Washington, DC  United States  20590
  • Performing Organizations:

    Upper Great Plains Transportation Institute

    1320 Albrecht Blvd.
    Fargo, ND  United States  58102
  • Principal Investigators:

    Lu, Pan

    Bridgelall, Raj

  • Start Date: 20160127
  • Expected Completion Date: 20180731
  • Actual Completion Date: 0
  • Source Data: MPC-505

Subject/Index Terms

Filing Info

  • Accession Number: 01601173
  • Record Type: Research project
  • Source Agency: Mountain-Plains Consortium
  • Contract Numbers: DTRT13-G-UTC38
  • Files: UTC, RiP
  • Created Date: May 31 2016 12:19PM