Investigation of derailment and wheel wear in a BEML metro coach under different operating conditions

Derailment and wheel wear in railways are of major concern which involve complex operating and conflicting dynamics parameters. Metro trains undergoes through sharp turns, steep gradients, frequent high acceleration and decelerations, and overloading during peak hours which heighten the multivariate ate aspect of the problems. In this work, an attempt has been made to investigate the derailment coefficient and wheel wear of all the eight wheels of a BEML (Bharat Earth Movers Limited) metro coach under different operating scenarios. Various running conditions are generated through response surface methodology (RSM) approach by varying vehicle speed, axle load and friction at the rail-wheel contact. For this, a multibody vehicle dynamics model replicating BEML metro coach is built in commercial software Simpack. The developed multibody dynamics model is validated from the field trials conducted in Kolkata, India, by matching vehicle motion and ride comfort indices along the track. Validated multibody dynamics model is then used for simulating different running scenarios according to the central composite design (CCD) scheme. Data generated from the multibody dynamics model under different operating scenarios are taken as inputs and outputs target data for a deep neural network (DNN) model. Results of the RSM approach indicate that lower friction at the rail-wheel contact is desirable for lower wear indices and smaller derailment coefficients. Operating speed, in the speed range considered, has little influence on wear index and derailment coefficient. Results of the developed DNN model demonstrate that the mean absolute percentage error (MAPE) value is lower than 4% for all the eight wheels in both training and test.