Multiobjective optimization based on polynomial chaos expansions in the design of inductive power transfer systems

Purpose - The purpose of this study is to decrease the computation time that the large number of simulations involved in a parametric sweep when the model is in a three-dimensional environment. Design/methodology/approach - In this paper, a new methodology combining the PCE and a controlled, elitist genetic algorithm is proposed to design IPT systems. The relationship between the quantities of interest (mutual inductance and ferrite volume) and structural parameters (ferrite dimensions) is expressed by a PCE metamodel. Then, two objective functions corresponding to mutual inductance and ferrite volume are defined. These are combined together to obtain optimal parameters with a trade-off between these outputs. Findings - According to the number of individuals and the generations defined in the optimization algorithm in this paper, it needs to calculate 20,000 times in a 3D environment, which is quite time-consuming. But for PCE metamodel of mutual inductance M, it requires at least 100 times of calculations. Afterward, the evaluation of M based on the PCE metamodel requires 1 or 2 s. So compared to a conventional optimization based on the 3D model, it is easier to get optimized results with this approach and it saves a lot of computation time. Originality/value - The multiobjective optimization based on PCEs could be helpful to perform the optimization when considering the system in a realistic 3D environment involving many parameters with low computation time.

Automated summary

The purpose of this study is to reduce the computation time in a three-dimensional environment by combining PCE and a controlled, elitist genetic algorithm. A PCE metamodel is used to express the relationship between quantities of interest and structural parameters, and two objective functions are defined for obtaining optimal parameters. The results show that this approach can achieve optimized results faster and save computation time compared to traditional 3D modeling optimization methods.