Estimation of rail-wheel contact forces and wear number using wheel tread temperature rise

Abstract

Temperature rise at the rail-wheel contact is directly correlated to the contact forces and wheel wear which are central to railway vehicle safety and stability. Temperature rise at the rail-wheel contact surface is influenced by the contact patch shape, size, location, contact forces and slip at the rail-wheel contact. In this work, a deep neural network (DNN) model is developed to predict the various contact parameters using temperature rise at the rail-wheel contact. Input parameters to the DNN model include operating (speed and acceleration), track (twist, elevation, gradient, curvature and superelevation) and temperature parameters (maximum temperature, full width half maxima and temperature peak location along the lateral direction), while, output to the DNN model include normal load, derailment coefficient, wear number and contact location. Data for training and test was obtained using a multibody vehicle dynamics model built in commercial software SIMPACK. The multibody vehicle dynamics model was validated using data from oscillation trials on a metro train operating on east-west Kolkata metro track. Wheel surface temperature rise is obtained using a two-step method. First, for each wheel, a two-dimensional boundary element model (BEM) is used to estimate heat partitioning at rail-wheel interface as a function of time. Next, frictional dissipation in contact patch along with heat partitioning information is fed into a three-dimensional BEM to obtain wheel surface temperatures. Multiple linear regression (MLR) and DNN models are then used to correlate temperature, track, and operating parameters to contact parameters. The developed DNN model accurately predict the contact parameters with R² greater than 90%. MLR model is only able to predict wear number accurately.