Affiliations: [a] National CIMS Engineering Research Centre, Tsinghua University, Beijing, China | [b] School of Engineering, University of Portsmouth, Portsmouth, UK | [c] Engineering Design Centre, Engineering Department, University of Cambridge, Cambridge, UK
Correspondence:
[*]
Corresponding author: Heming Zhang, National CIMS Engineering Research Centre, Tsinghua University, Beijing 100084, China. E-mail: [email protected].
Abstract: Multidisciplinary collaborative simulation is an important technique for the development of complex engineering systems such as aircrafts and automobiles. It emphasizes the synergic collaboration of multidisciplinary computational models, involving the exchange of simulation data generated in parallel at runtime from the numerical integration processes of these models. Current research on the interaction of models in multidisciplinary collaborative simulation is mainly focused on performing data exchange at fixed intervals, i.e. macro time steps, and using interpolation/extrapolation to find out values at the points in time when data are not available. However, the selection of an appropriate size for macro time tends to be complex and iterative. Firstly, it is mostly based on experience rather than the estimation of simulation errors by analyzing the numerical integration processes involved. Moreover, it is hardly possible to select a fixed step size suitable for all the stages of the simulation process. This paper presents a novel variable-step algorithm which is able to adjust the step size automatically at simulation runtime based on the evaluation of truncation errors. This algorithm not only avoids the instability and inaccuracy caused by a fixed large step but also speeds up the simulation by using a smaller step when a large one is not necessary. It is demonstrated in the numerical experiments that the variable-step algorithm can achieve improved simulation performance in terms of both speed and accuracy.
