Fault detection and classification method to prevent derailment of rolling stock

This paper proposes a method to detect and classify faults that lead to the derailment of rolling stock. The proposed approach involves the measurement of parameters related to the bearing, wheel and track defect of a rolling stock. It suggests a simple method of classification of faults based on elimination using a fuzzy logic based fault classification system. Rolling element bearing temperature and its rate of change, difference in temperature between a normal bearing and bearing under consideration, vibration of the wheel under consideration and vibrations data from other wheels are used to analyze and classify the faults. The system suggested is demonstrated with simulation using MATLAB. In an effort to classify the faults of rolling stock that lead to derailment, the vibration data from the bearing of an electric motor is chosen in place of vibration data from the bearing of rolling stock. The vibration signals of the electric motor is obtained from bearing data centre of Case Western Reserve University with kurtosis chosen as the parameter to be obtained for fault classification. The simulation values for temperature and rate of temperature is obtained by considering temperature as a function of vibration and then extrapolating the results from an experiment on bearing temperature of another machine. Also simulation is carried with time of high value of RMS as one of the parameters. A fuzzy logic based classification is done. The method suggests that classification of faults based on elimination can be effectively done and can be the basis to analyze faults of rolling stock with real time signals.