Keywords: anomaly detection, derailment coefficient, statistical approach, Nadal’s formula, lubrication condition, inner-rail L/V, bogie faults, L/V monitoring system, condition monitoring, correlation.

Monitoring the running condition of railway vehicles in curves and detecting abnormalities in bogies are essential for ensuring safe and stable railway operations. The derailment coefficient is widely used as a fundamental indicator for evaluating vehicle running safety. It is conveniently obtained as the ratio of lateral force to wheel load (L/V) acting on the outer rail. However, the derailment coefficient is influenced not only by vehicle performance but also by various factors such as track conditions (e.g., curve radius) and running conditions. Therefore, it is difficult to detect bogie abnormalities based solely on the value of the derailment coefficient. On the other hand, the derailment coefficient on the outer rail (outer L/V) exhibits a strong correlation with that on the inner rail (inner L/V). Based on this correlation, it is hypothesized that normal conditions correspond to maintaining this relationship, whereas deviations indicate abnormal conditions. This paper proposes a statistical approach for anomaly detection of derailment coefficients considering the inner rail L/V. By constructing a scatter plot of inner L/V versus outer L/V using a dataset comprising 4,518 measurements measured in revenue service, and defining ±3? and ±5? boundaries based on deviations from a regression line, anomalous conditions can be mechanically identified. The results demonstrate that the proposed method can detect bogie faults, such as wheel load imbalance, that cannot be identified from the derailment coefficient alone.

download the full-text of this paper (PDF, 13 pages, 1173 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description