Rail Vehicle Derailment Detection Using Acceleration Quantile Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting rail vehicle derailments are limited in their ability to distinguish between derailments and other events causing high accelerations, such as driving over rail joints or flat spots, leading to false alarms and delayed detection.
Innovation Solution
The method involves determining the p-quantile of the acceleration signal's time profile within a specified time window and comparing it with a threshold value, allowing for differentiation between derailments and other events, enabling reliable detection even after the event has occurred.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If acceleration signal is monitored continuously to detect derailments, then derailment detection capability is improved, but false alarms increase due to other events causing high accelerations
Solution Approach 1:
The patent applies local quality by analyzing specific characteristics of acceleration signals rather than treating all high acceleration events uniformly. It identifies and weights different signal patterns differently, giving minimal weight to outliers caused by rail joints and switches while maintaining sensitivity to derailment patterns. This selective analysis approach reduces false alarms while preserving derailment detection capability.
Solution Approach 2:
The patent changes parameters by introducing a time window parameter and a weighting parameter for different signal characteristics. By adjusting these parameters, the system can distinguish between transient high accelerations (false alarms) and sustained derailment patterns. The ability to modify analysis parameters allows the system to adapt to different operating conditions while maintaining reliable detection.
2Device complexity
If traditional acceleration threshold methods are used, then simple detection is achieved, but differentiation between derailments and other events is poor
Solution Approach 1:
The patent introduces dynamics by implementing a time window that moves with the rail vehicle, allowing the analysis period to adapt to the vehicle's speed and position. This dynamic approach enables the system to maintain appropriate analysis duration regardless of operating conditions, improving event differentiation without requiring complex fixed-threshold systems.
Solution Approach 2:
The patent uses the time window as an intermediary between the raw acceleration signal and the final detection decision. This intermediary structure allows for weighted analysis of signal characteristics within the window, enabling precise differentiation between event types while keeping the overall system architecture relatively simple.
3Loss of time
If derailment detection is performed only at the actual time of derailment, then immediate detection is achieved, but post-derailment detection capability is lost
Solution Approach 1:
The patent applies preliminary action by continuously monitoring and storing acceleration signal characteristics within a time window before, during, and after potential derailment events. This continuous preliminary analysis ensures that derailments are detected at their actual occurrence while also maintaining the capability to identify and confirm derailments that occurred slightly earlier, preventing loss of detection capability.
Data Source
Figure 1~2
Figure 3~4
AI summary
The invention relates to a method for recognising when at least one wheelset of a rail vehicle derails, in which method an acceleration signal which represents the acceleration of the wheelset in one direction is measured during travel, the p-quantile is determined for a number of values of the time curve of the acceleration signal within a predetermined time window, p corresponding to a predetermined value higher than 0 and lower than 1, the p-quantile is compared with a predetermined threshold value and checked in order to determine whether the p-quantile exceeds or falls below the predetermined threshold value.