Application of automatic classifiers for condition monitoring of railway rolling stock

The evolution of technology towards the automation of industrial processes and the advances in interconnectivity have given way to what is known today as industry 4.0. These advances are of particular interest in the area of predictive maintenance of machines, where machine learning techniques have considerably improved condition diagnosis of machinery. This is of special importance in the railway industry, where maintenance constitutes an important part of its operating costs. This paper studies the application of machine learning techniques to vibration signals originating from a railway axle, tested on a railway test bench, through support vector machine algorithms for fault detection. A feature selection scheme composed of a series of sensitivity analyses is proposed in order to determine the best signal features for classification. The subsequent hyperparameter optimization proposed consists of a series of sensitivity analyses in order to determine the values of each parameter that result in a classifier with the most accuracy. Lastly, the effect of the location of the sensors in the axle from which the vibration signals are obtained is studied in order to determine their most apt configuration.

Application of automatic classifiers for condition monitoring of railway rolling stock Articles