Possibilities of failure detection and identification (FDI) in a train localization system

Traditional train control has been discrete speed and position control via fixed block quantization. The principle of continuous train control especially its high performance version, moving block, is being used more for high speed trains or for urban transit, introducing the problem of obtaining vital continuous measurements of position and speed. Omitting position discrete measuring devices (except for synchronization), there is the problem of how to obtain a safety related interval of confidence, the true value of the measured quantity is guaranteed to stay within according to the safety standard. It has been shown that use of consistent data is an integral aspect to achieve this objective. Therefore FDI algorithms have to be applied to the measurement data. There are several requirements on the FDI algorithms, if used in vital systems. These requirements are formulated and it is examined how FDI algorithms fulfil these requirements. Considering the aspect of safety issues the authors further propose a classification scheme for these algorithms focussing especially on analytical redundancy. The goal of this scheme is firstly to allow an orientation in the `jungle of FDI´ and secondly to provide the possibility to project results for one algorithm of a class to an other algorithm of that particular class. The classification scheme as well as the examinations of the algorithms´ ability to meet the safety requirements are explained by applying examples to measurement data obtained from a complex train simulator