Development of a simulation backplane with dynamic configurability

The subject of this thesis was the development a simulation backplane for coupling an electrical simulator with a mechanical finite-element simulator although the backplane is applicable to any simulator with the proper architecture. To tradeoff performance and accuracy during the analysis, a dynamic configuration option was included, where different simulators and models in a simulator can be switched in and out during the analysis. This option provided the designer with the flexibility to analyze all details of simulations from different modeling representations to optimizing the simulation performance within the same analysis. This backplane was able to transverse multiple coupling architectures to be a configuration tool for simulating hybrid environments with dynamic changes. In this work, the dynamic configurability procedure was outlined with other issues and procedures for coupling multiple simulators. To improve upon the basic coupling process, different interface configurations were examined in different casualty-based formats. Specifically, conventional interface combinations were compared under different sensitivity calculation criteria and methods to find the interface or combination with the best iteration efficiency and convergence. The iteration efficiency was typically determined by the sensitivity calculation options while the convergence was determined by the interface combination between the simulators and by the backplane initialization sequence. The optimum convergence for any conventional interface combination was 93%. From these analyses, a dynamic-interface configuration procedure was developed based on coupling conditions and variable causality to identify the interface configuration with the best chances of convergence. A tiered dynamic interface procedure had convergence of 95% and equivalent iteration efficiency with the conventional interfaces. Finally, a flow correction method using behavioral models was examined, where a predictor and corrector process was implemented that allowed more versatility in the coupling process and improved initialization compared to the other interfaces. The flow correction process had a convergence of 93% tested over behavioral models with different accuracy constraints. A sensitivity calculation problem limited the success of the flow correction procedure and caused the iteration inefficiency to be two times larger than the other procedures.