RAILWAY VEHICLE MAINTENANCE SYSTEM WITH MODULAR RECURRENT NEURAL NETWORKS FOR PERFORMING TIME SERIES PREDICTION

摘要

The present invention relates to a method and apparatus for providing maintenance of a rail vehicle (10) by time series prediction, wherein at least two recurrent neural network models (200-1 to 200-m) which correspond to predetermined parts of the rail vehicle (10), that are exposed to degradation, are constructed, wherein the models (200-1 to 200-m) are used for time series prediction of sensor output information which relates to a corresponding part of the rail vehicle (10) within a condition based maintenance system of the rail vehicle (10), and wherein a maintenance output (P1 to Pm) is generated for each of the predetermined parts based on the time series prediction.