| Computer simulation of physical systems has been widely adopted in many science and engineering disciplines. Due to the increasing affordability of high computation power, it is now possible to simulate complex mechanical systems with very high accuracy. As a simulated system grows in size and complexity, it becomes necessary to model various subsystems independently by different engineers at different times. This type of modular system modeling allows partial upgrades or model changes in a large complex system so that other parts of the system model can be reused in a time and cost-effective manner. The subsystem models, or submodels, are sometimes written in different languages due to different strengths and capabilities of various software packages. In order to overcome the difficulty with simulating dynamic systems with multiple submodels written in different languages, co-simulation is currently being studied by a growing number of researchers. Co-simulation is a simulation methodology whereby multiple submodels developed in different software are simulated together in a single framework.;In this research, the feasibility of co-simulation of systems with multiple submodels is studied in two different simulation configurations: serial and parallel. In order to study the accuracy of the different simulation configurations, the explicit Euler method is applied to the simulation of a simple two degree-of-freedom mass-spring system in three different simulation configurations: integrated, serial, and parallel. Then, the results are compared with the analytical solutions to study simulation accuracies. The results show that the parallel simulation shows the same results as the integrated system model while the serial simulation exhibits larger errors than the other two configurations.;As a practical application of co-simulation, a full vehicle system is co-simulated for the full FTP cycle in the serial and parallel simulation configurations. The vehicle model consists of a high fidelity GT-POWER engine model and a medium fidelity vehicle model developed in MATLAB/Simulink. The simulation results for the vehicle model showed that the parallel co-simulation could reduce the computation time by 25.7% for this particular vehicle model, while the performance variables showed that reasonable accuracy could be achieved. |
