Measurement data treatment in multi-sensor applications for railway vehicle inspection

Achievement of high safety for railways comprises different aspects. Beside train control technology, also the examination of fault states of driving trains makes a contribution to safety assurance. The past inspection through human train station inspectors is going to be replaced by technical equivalents. Available detection systems focus only on particular fault states. Reliability and accuracy of such systems often suffer from heavy measurement conditions (e.g. high train speeds, disturbing environmental influences). Thus, a new approach consists in the usage of data of different sensor systems for evaluating train fault states from a more global point of view ("Checkpoint"). It implies the necessity of data conjunction for gaining more reliable results. In this work a concept for uniform data treatment is presented. It is based on sensor system independent data representation and on application related data storage for easy and flexible data conjunction. The uniform data representation also includes error descriptions of the values. Furthermore, problems to estimate fault state recognition errors are pointed out and a feasible approach is outlined.