CAJ | 학술논문

提出一种基于粗糙集和遗传算法的BP神经网络故障诊断方法,解决基本BP网络收敛速度慢、精度低、易陷入局部极小值问题.运用粗糙集理论对训练样本进行属性约简,简化BP网络输入维数.设计2次遗传算法训练BP网络,第一次优化神经网络隐含层节点个数,第二次在神经网络结构确定的情况下,优化网络连接权值.以柴油机进、排气阀故障为例,应用提出的方法进行仿真,仿真结果证明了该方法能够优化神经网络结构,提高故障诊断速度和准确率.
제출일충기우조조집화유전산법적BP신경망락고장진단방법,해결기본BP망락수렴속도만、정도저、역함입국부겁소치문제.운용조조집이론대훈련양본진행속성약간,간화BP망락수입유수.설계2차유전산법훈련BP망락,제일차우화신경망락은함층절점개수,제이차재신경망락결구학정적정황하,우화망락련접권치.이시유궤진、배기벌고장위례,응용제출적방법진행방진,방진결과증명료해방법능구우화신경망락결구,제고고장진단속도화준학솔.
A BP network faults diagnosis method based on the rough set and genetic algorithm is presented,solving problems of low precision,slow constringency and local minimum. In order to simplify the input of BP network,the attributes of training samples are reduced by the rough set theory. Double genetic algorithms are designed to train BP network,for the first time,to optimize the number of hidden layer neurons,secondly to optimize the weights of network. The faults diagnosis for the diesel intake valve and outlet valve is simulated using the proposed method. The results show that the proposed method can optimize the structure of neural network and improve the rate and precision of fault diagnosis.