The Reduction of Simulation Software Execution Time for Models of Integrated Electric Propulsion Systems through Partitioning and Distribution

Abstract

Software time-domain simulation models are useful to the naval engineering community both for the system design of future vessels and for the in-service support of existing vessels. For future platforms, the existence of a model of the vessel's electrical power system provides a means of assessing the performance of the system against defined requirements. This could be at the stage of requirements definition, bid assessment or any subsequent stage in the design process. For in-service support of existing platforms, the existence of a model of the vessel's electrical power system provides a means of assessing the possible cause and effect of operational defects reported by ship's staff, or of assessing the possible future implications of some change in the equipment line-up or operating conditions for the vessel. Detailed high fidelity time-domain simulation of systems, however, can be problematic due to extended execution time. This arises from the model's mathematically stiff nature: models of Integrated Electric Propulsion systems can also require significant computational resource. A conventional time-domain software simulation model is only able to utilize a single computer processor at any one time. The duration of time required to obtain results from a software model could be significantly reduced if more computer processors were utilized simultaneously. This paper details the development of a distributed simulation environment. This environment provides a mechanism for partitioning a time-domain software simulation model and running it on a cluster of computer processors. The number of processors utilized in the cluster ranges between four and sixteen nodes. The benefit of this approach is that reductions in simulation duration are achievable by an appropriate choice of model partitioning. From an engineering perspective, any net timing reduction translates to an increase in the availability of data, from which more efficient analysis and design follows.