Parallel computing of wheel-rail contact

Abstract

Railway wheel–rail contact simulations are the most important and time-consuming tasks when simulating the system dynamics of vehicles. Parallel computing is a good approach for improving the numerical computing speed. This paper reports the advances in parallel computing of the wheel–rail contact simulations. The proposed method uses OpenMP to parallelise the multiple contact points of all the wheel–rail interfaces of a locomotive model. The method has been implemented in the vehicle system dynamics simulation package GENSYS. Simulations were conducted using two numerical solvers (4th Runge-Kutta and HeunC) and a maximum of four computer cores. Simulation cases have shown exactly the same numerical results using serial computing and parallel computing, which prove the effectiveness of the parallel computing method. The HeunC solver achieved the same simulation results and is 3.5 times faster than the 4th Runge-Kutta method. Simulation results obtained from both numerical solvers show that parallel computing using 2, 3 and 4 computer cores can improve the simulation speeds by roughly 29, 39 and 41%, respectively. There is an apparent diminishing of the rate of improvement due to the increase of the communication resource overhead when more computer cores are used. Using up to four computer cores does not require revision of the GENSYS code, and simulations can be executed using personal computers.