Enhanced digital model based on Kalman filter for fault detection in railway rolling stock

Nowadays, the integration of the different pillars on which the industry 4.0 or 5.0 paradigm is based enables for the digitisation of all the tasks related to the life cycle of a product or service. Nevertheless, great efforts have been made in the organisational and decision-making areas. In this sense, the maintenance tasks, and in particular the activities related to the analysis of the operating status of the elements of a mechanical system, must be reformulated and oriented towards digitisation. To achieve this goal with success, the development of models for condition monitoring is a novel topic of rising interest both in industry and academia, models oriented to the formulation of a Digital Twin (DT). In particular, the implementation of such systems in the railway industry has the potential to significantly reduce maintenance costs for the rail as well as rolling stock and predict their lifespan. When it comes to rolling stock, its dynamic nature makes it necessary to consider both the dynamics in short time spans as well as the variation of parameters at longer timescales. In this manuscript, a new model to estimate failures in the rolling stock system is presented. This method involves the use of techniques which lead to the development of a digital twin (DT) of this complex mechanical system. The system oriented to DT comprises a dynamical model of distributed mass and a Kalman filter to predict the dynamical behaviour of the system based on the information obtained with real data from a test bench. The results prove the soundness of the research as the model follows the real behaviour in time. Also, in order to validate the use of the model for condition monitoring purposes a proposal to detect cracks in railway axles is also presented.