Condition-Based Maintenance for Normal Behaviour Characterisation of Railway Car-Body Acceleration Applying Neural Networks

Recently, passenger comfort and user experience are becoming increasingly relevant for the railway operators and, therefore, manufacturers as well. The main reason this to happen is that a clear differential value considered by passengers final customers. Passengers’ directly related accelerations received through car-body of train. For reason, suspension damping components must be maintained in perfect condition, assuring high levels quality. An early detection any potential failure these systems derives better maintenance inspections’ planification more sustainable approach whole train strategy. In paper, an optimized model based on neural networks trained order predict lateral accelerations. Comparing predictions values measured train, normal characterisation dynamic behaviour can determined. Any deviation from will imply loss or degradation components. This has been with dataset specific unit, containing variables recorded every second during year 2017, including vertical accelerations, among others. A minimum average error 0.034 m/s2 obtained prediction means approximately 2.27% typical maximum estimated vehicle body reflected EN14363 coaches (1.5 m/s2). Thus, successful achieved. addition, evaluated real situation which noticed lack comfort, achieving excellent results atypical Therefore, it possible measure acceleration deviations standard causing passengers, alert sent operator maintainer non-programmed intervention at depot (predictive maintenance) board (prescriptive maintenance). As result, condition-based (CBM) methodology proposed avoid could end complaints speed limitation due safety reasons excessive acceleration. highlights concept energy efficiency