Abnormality detection via SVDD technique of motor-generator system in HEV

Abstract

This paper introduces a method to detect abnormality of MGS (Motor-Generator System) in HEV (Hybrid Electric Vehicle) using its temperature. The MGS in HEV consists of two Motor-Generators (MG1, MG2), Compound Gear Unit, and etc. The MG1 is to act as a generator in conventional internal combustion engine. And the MG2 is an electric motor to rotate wheel of vehicle using saved electricity in battery or using produced electricity via the MG1. In case of overheating, the electric motors are easily damaged because resistance of wires in motor is abnormally changed. Therefore, detection of abnormally changed temperature in motors (MG1 and MG2) is essential. In this study, the temperature distribution of two Motor-Generators is observed simultaneously in 2-dimensional space. A boundary region of normal operation temperature of two motors is obtained via SVDD technique utilizing Gaussian kernel, one of the most widely being used Mercer kernels. Linear SVDD technique generates boundary of exact ball shape, however SVDD technique using Gaussian kernel can generate nonlinear boundary of distorted ball shape. Abnormality boundary comparison is made between the obtained boundary via SVDD technique and those obtained from conventional temperature range checking method. In order to compare the performance of proposed method, the actual vehicle operation data in excessive driving condition on mountain road is adopted. In verification, simulation shows that warning time due to proposed method is faster and more efficient than those due to conventional method. It is also shown that the reliability of the Motor-Generator System can be improved by using the proposed abnormality detection method.