Abstract
Uneven wheel loads due to wagon defects have in the past been identified as causal factors in freight wagon derailments leading to Rail Accident Investigation Branch (RAIB) recommendations to industry to address the cause, effects and monitoring of defect related uneven wheel loads.
Wheel Impact Load Detectors (WILD) are installed on Network Rail infrastructure and routinely measure wheel loads from passing traffic, primarily to detect wheel defects such as flats. However, work over recent years has demonstrated that the WILD data can also give information relating to the vehicle and its load, specifically vehicle or load imbalance. The use of the WILD data for other aspects of traffic monitoring, such as identifying vehicle defects like wagon frame twist, is theoretically possible but as yet unproven.
This project aims to determine the viability of using WILD data for wagon defect detection. Such defects could include a twisted vehicle frame, twisted bogie frame and abnormal suspension behaviour or condition.
The change in wheel load distribution for a number of defect types was calculated using simplified mathematical models of the vehicle suspension and defect characteristics. The relationship between common metrics and wheel unloading (a reliable derailment indicator) were defined. This part concluded that the diagonal wheel load imbalance is an appropriate metric for detecting defective vehicles which could have reduced derailment resistance.
A statistical process for identifying meaningful trends in wheel loads for a specific vehicle over a period of time has been developed and applied to a 12-month WILD dataset of RFID tagged (identifiable) vehicles. The process included the characterisation of the data set to ensure statistically meaningful observations.
From this process a shortlist of 28 vehicles was created for further investigation with support from the vehicle owners/maintainers. In several cases, step changes observed in the performance of the vehicle were correlated with recent maintenance activities. Wheel unloading estimations for a number of cases were similar to those calculated by RAIB as being contributory factors in historic derailments.
Based on initial feedback from the vehicle maintainers the project concluded that it is feasible to use WILD data for vehicle defect detection by applying a statistical analysis process. The acquisition of further information regarding vehicle condition is ongoing and more detailed correlations will be possible in the future when the data becomes available.
It was recommended that the feasibility of implementing a statistical analysis process in the WILD data stream should be assessed, with the aim of providing relevant stakeholders with offline information to inform maintenance scheduling and inspections where necessary on an ongoing basis.
Wheel Impact Load Detectors (WILD) are installed on Network Rail infrastructure and routinely measure wheel loads from passing traffic, primarily to detect wheel defects such as flats. However, work over recent years has demonstrated that the WILD data can also give information relating to the vehicle and its load, specifically vehicle or load imbalance. The use of the WILD data for other aspects of traffic monitoring, such as identifying vehicle defects like wagon frame twist, is theoretically possible but as yet unproven.
This project aims to determine the viability of using WILD data for wagon defect detection. Such defects could include a twisted vehicle frame, twisted bogie frame and abnormal suspension behaviour or condition.
The change in wheel load distribution for a number of defect types was calculated using simplified mathematical models of the vehicle suspension and defect characteristics. The relationship between common metrics and wheel unloading (a reliable derailment indicator) were defined. This part concluded that the diagonal wheel load imbalance is an appropriate metric for detecting defective vehicles which could have reduced derailment resistance.
A statistical process for identifying meaningful trends in wheel loads for a specific vehicle over a period of time has been developed and applied to a 12-month WILD dataset of RFID tagged (identifiable) vehicles. The process included the characterisation of the data set to ensure statistically meaningful observations.
From this process a shortlist of 28 vehicles was created for further investigation with support from the vehicle owners/maintainers. In several cases, step changes observed in the performance of the vehicle were correlated with recent maintenance activities. Wheel unloading estimations for a number of cases were similar to those calculated by RAIB as being contributory factors in historic derailments.
Based on initial feedback from the vehicle maintainers the project concluded that it is feasible to use WILD data for vehicle defect detection by applying a statistical analysis process. The acquisition of further information regarding vehicle condition is ongoing and more detailed correlations will be possible in the future when the data becomes available.
It was recommended that the feasibility of implementing a statistical analysis process in the WILD data stream should be assessed, with the aim of providing relevant stakeholders with offline information to inform maintenance scheduling and inspections where necessary on an ongoing basis.
| Original language | English |
|---|---|
| Publisher | University of Huddersfield |
| Commissioning body | Rail Safety and Standards Board |
| Number of pages | 61 |
| Volume | IRR Ref: 110/210 |
| Edition | 2 |
| Publication status | Published - 25 Jan 2021 |
