New opportunities in an old industry Power Products – challenges in field analysis and optimization
Simulation of multi-physics phenomena in power devices is becoming crucial for power industry. Industry demands for energy efficiency and cost savings make optimization a logical next step after field analysis. Challenges of numerical technologies from this industrial perspective are presented hereafter.
Once a simulation algorithm has been developed to the stage of an industrial process, the next natural step is to pursue semi-automated simulations (possibly embedded in a design system), for which the target user is no longer an expert in numerical technologies.
The optimization (based on 3D simulations) of power devices in industry is in its incipient phase, since problems are generally multi-objective, governed by a relatively high number of parameters, and the objective functions are difficult to evaluate. In order to address these pitfalls, one approach is to come up with an automated analysis process empowering a designer to choose among several sub-optimal solutions. The realization we propose involves an application driving the field solver over the memory bus and exchanging information via files with an optimization software package, in this particular case modeFRONTIER. A set of parameters is transmitted from the optimization tool to the field solver, and results of the simulation necessary for the objective function evaluation are sent back to the optimizer. Two optimization runs were performed, one based on a multi objective genetic algorithm (MOGA) implementation and one on a Non-dominated Sorting Genetic Algorithm II (NSGA-II).