Model-Based Estimation of Wheel/Rail- and Brake Disc/Pad-Friction for Railway Vehicles

Abstract

This work presents a novel application of model-based online estimators for the simultaneous identification of wheel-rail adhesion and brake pad-disc friction for railway vehicles. The synthesis of an estimator for these two parameters is necessary, since their direct measurement in daily operation is not feasible due to technical and economic reasons. The detailed knowledge of adhesion and friction allows for a significant improvement of capacity and reliability of the existing rail network, e.g. by contributing to continuous track and vehicle monitoring. In order to exploit this potential as quickly as possible in daily operation, a ready to use design is required, which is challenging for the estimator synthesis in two ways. Firstly, it demands for a minimum set of mostly in-service sensors and, secondly, an estimator framework has to be generated that suits the limited computational resources of regular train control units. Further challenges arise from the simultaneous estimation of two partly interconnected parameters and their identification for every single wheelset. To meet these demands, the present work combines a detailed analysis of railway vehicles with a control theoretic estimator synthesis. Based on the analysis of railway vehicles, two nonlinear systems are generated and their observability is verified. To complement this qualitative observability evaluation, a quantitative and signal-based correlation measure is developed. This novel correlation measure considers sensor properties, e.g. signal-to-noise ratio. Besides the in use measurements of wheelset and train velocity, the two most promising measurements according to the signal-based correlation are longitudinal acceleration detected at the car body and vertical deviation in the primary suspensions. The two nonlinear system variants are combined with two estimator concepts, Extended and Unscented Kalman Filter, so that four configurations are applied to test rig data. The 50 tested scenarios comprise different variations, which ensure robustness, reliability, and accuracy of the estimators for a broad spectrum of operating conditions. To confirm the test rig results, the estimation environment is adapted to track test data. With the given measurement equipment in the test vehicle the plausibility of estimation results could be verified. Finally, two adhesion-based control concepts are introduced that address repeatability of brake distances and their reduction, respectively. These brake control concepts demonstrate one possible application in the wide field, which is opened by the reliable identification of wheel-rail adhesion and brake pad-disc friction.