A novel simple and effective class discriminator of motor quality based on genetic algorithm

• Main Authors: Pao-Shuang Chen, 陳葆霜
• Other Authors: 廖炯州
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/s9sdz9

碩士 === 健行科技大學 === 電子工程系碩士班 === 106 === This paper proposes a simple and effective class identifier algorithm for motor quality based on the genetic algorithm. This algorithm is composed of the following three parts, namely (a) the pre-processing stage of signals: its main purpose is to extract and amplify motor current signals and remove noises; (b) the selection stage of main feature points: the main purpose is to select the most important feature points that can express the characteristics of the original signal from a large number of original feature points. This stage can simplify calculation of the large amount of data needed to perform the classification work; (c) the class identification stage of motor quality: the class identification method of DC motor quality utilizes the genetic algorithm. The used genetic algorithm includes five parts: encoding and decoding, fitness evaluation, crossover, mutation, and reproduction. In short, the genetic algorithm is used to find the sub-optimal solution by encoding and decoding with repeated and continuous operations until the closest one to the best solution is found. Four motor quality classes are identified, including the normal motor quality type (Type-good), and three fault motor quality types (Error type-1, Error type-2, and Error type-3). Experimental results show that the Type-good classification accuracy rate of normal motors is 95.55%, and Error type-1, Error type-2, and Error type-3 classification accuracy rates of faulty motors are 86.66%, 82.35%, and 86.95%, respectively. The total classification accuracy rate is about 90.00%.