Statistical Process Control for Fault Identification of a Rolling Stock

Predictive maintenance and quality control are quite related areas, as they both are focused on monitoring and improving systems and processes performance. In the last decades, several proposals of integrating a Statistical Process Control approach to modeling and solving reliability, safety and maintenance issues have been presented in the scientific literature. In the present paper we describe our experience with the development of a methodology to carry out an automated diagnostic process based on the control charts application for a rolling stock. The operational data are available in real time thanks to an on board diagnostic system, able to send the information to a ground-fixed server but the huge number of data and their complex structure are not helpful to perform a real preventive maintenance strategy. So the aim of the project was the development of an approach able to improve the maintenance performance. After a first phase of data analysis, treatment and filtering; in order to identify the best control charts for this application, several cc types have been applied to a set of data that come from limited and well-known time intervals. The results have been compared by the information available in the CMMS for the maintenance activity and a new methodology for data valorization has been suggested.