Effective Data-driven Method Based on Bayesian Approach for Performance Estimation of Wound-field Motors

Abstract

This study proposes a novel data-driven method based on Bayesian approach. A d/q axis flux map and the non-linear torque characteristics of wound-field motors were estimated using a single layer neural network to reduce the number of finite element analysis and load tests. Moreover, to improve estimation accuracy even with a small number of learning data, training data-set to be input into machine learning were selected based on a Bayesian approach. The proposed method improves the estimation accuracy compared to that of the conventional data-driven method. The proposed method is validated by applying it to the numerical and experimental problem. Moreover, estimated results are compared with those of the conventional methods.