Co-simulation methods for train braking dynamics

Abstract

The development of train dynamics models usually started with simplified empirical air brake models. Fluid dynamic air brake models are sometimes required for train dynamics simulations to utilise their advantages in physical descriptions and model scalabilities. Due to the high complexity of fluid dynamic air brake models, the most economic approach to add such models to train dynamics simulation is via co-simulations. This paper presents three methods that can be used for co-simulations of train braking dynamics. These models are based on Transmission Control Protocol/Internet Protocol (TCP/IP), Open Multi-Processing (OpenMP), and Message Passing Interface (MPI) techniques, respectively. Requirements for co-simulations of train braking dynamics are discussed. Code structures regarding how to implement these techniques for train braking dynamics simulations are presented and discussed. Demonstrative simulations using dummy programs are presented to show the effectiveness of these methods. Comparatively, OpenMP, MPI, and TCP/IP have an increasing level of complexity for implementations as well as commination time overhead. OpenMP and MPI also allow parallel computing to increase computing speeds. For the case of train braking dynamics simulations, the authors’ recommendation regarding the optimum application sequence is OpenMP, MPI, and TCP/IP.