Robust on-board detection of hunting on railway vehicles via dynamics-based multi-sensor stochastic methods

This study focuses on the introduction and comparative assessment of multi-sensor parametric stochastic methods employing bogie lateral acceleration signals for on-board detection of hunting initiation in operating railway vehicles. Two novel methods, a Vector Degree-of-Stochasticity (V-DS) based and Vector Damping Ratio (V-DR) based, are introduced as proper extensions of our recent single-sensor counterparts, with the main focus being on potential improvements in detection performance and robustness to suspension faults, worn wheels, worn tracks, and 'unseen' operating scenarios. The performance of both methods is systematically studied via Monte Carlo simulations with an experimentally validated numerical model and certain field tests. The results indicate impressive performance, surpassing that of the best single-sensor counterparts and alternative non-parametric multi-sensor methods by achieving mean Detection Delay Time reduction by an order of magnitude, excellent detection characteristics, and superior robustness. The postulated methods thus constitute promising solutions for effective on-board detection of hunting initiation on railway vehicles.