Condition-based maintenance using roadside monitoring systems

Maximize the use of current roadside monitoring systems for CBM-related issues.

The detection of the degradation of rolling stocks in early stages is important to avoid stoppages, immobilization of wagons and delays that are costly and highly inconvenient, as they dramatically reduce the capacity of the infrastructure and the availability of vehicles. As a result, it is primordial to assess the degradation of vehicles using appropriate condition monitoring techniques. The condition monitoring (CM) applied in railway systems is divided into two categories: wayside monitoring systems and on-board monitoring systems. In general, for defects developing rapidly or for defects not possible to detect by the wayside equipment, on board systems are appropriate. For slowly evolving defects, wayside monitoring has the advantage of not requiring large scale installations of measurement equipment in many vehicles. In general, wayside measurements systems have traditionally been focused on detecting critical events which could result in additional failures, damages to the infrastructure or disturbances to the traffic.

However, to increase the usability of the current wayside monitoring system, new strategies should be developed to extract information related to the wheelsets, not only for safety related issues but as well for improving the condition monitoring in terms of maintenance perspectives for the vehicle owners. As a result, methods to extract relevant information have been developed within the FR8RAIL framework. A deeper analysis of the Hot-Box/Hot-Wheels and Wheel Impact Detectors (WILD) has been performed on a specific railway line from Aitik to Skellefteå in the north of Sweden. The study focuses on several key points that may improve the condition monitoring for wheelsets:

• Ensure the quality of transferred data between the different detectors, since deviations can occur at different locations (variation between HB types, position along the line, load influence, seasonality, dynamic properties…). These deviations could give raise to false alarms, as well as affecting preselected features used for prognostics
• Development of a concept (Figure 1) based on extracting temperature signatures of trains or wagons from the Hot-Box/Hot-Wheel measurements. This concept could help detecting anomalies related to a change of condition (lubrication, dirt, misalignment or unknown reasons) without reaching the current safety limits
• Choose appropriate feature according to the wagon type for the Wheel Impact Detectors. Train dynamics, load and speed may influence the readings of these specific features
• Investigate - from the available data - specific wheels or wagons that have slower degrading processes, which can be tracked, based on the dynamic load as the main feature (Figure 2). The degradation model may depend on the failure mode (stepwise, linear, exponential…). A labelling of failure mode and maintenance for specific wheels could be of interest for developing degradation models using wayside monitoring

By combining the quality check of various detectors, a rescaling of the features as function of specific predictors, and tracking down well-chosen features from RFID-tagged wagons as function of time or running distance, the possibilities to transition towards a pro-active approach may be feasible, especially for Wheel Impact Load detectors. To ensure that the case studies showing a slow degrading process are related with the wheelsets degradation, the infrastructure manager should be in close relation with the vehicles owners in order to label the current data set in terms of failure modes and maintenance activities. This interaction would lead to more accurate degradation models, which may result in a better estimate of the remaining useful life.

Sponsor: FR8RAIL – Trafikverket
Researchers: Matti Rantatalo (PL), Florian Thiery