On chassis predictive maintenance and service solutions: An unsupervised machine learning approach

Abstract

Predictive maintenance is a key component in cost reduction in automotive industry and is of great importance. Besides cutting the costs, predictive maintenance can improve feeling of comfort and safety, by early detection, isolation and prediction of prospective failures. That is why automotive industry and fleet managers are turning to predictive analytics to maintain a lead position in industry. In order to predict and mitigate components failure in advance, measurements data from vehicle parts are collected from sensory system mounted on vehicle parts, and employed to evaluate the health condition of the components. An unsupervised learning solution is proposed, in this work, for automatic processing, diagnosis and isolation of faults. This solution is used for advanced analytics of the collected time-series data in the back-end (assuming that faults were reported based on spectral analysis). A literature study on maintenance types with emphasis on predictive maintenance with application to chassis failure is done. Chassis failures and conventional failure detection techniques are also covered in this study. A data acquisition was done at Hällared test track and labelled data regarding error states of interest were collected. Performance of the proposed method was validated by automatics clustering of two error states, i.e. low tyre pressure and faulty wheel hub.