基于Chwa&Hakimi模型的GA-BPFD算法

GA-BPFD algorithm based on Chwa & Hakimi model

提出一种Chwa&Hakimi模型下基于遗传算法优化的BPFD算法——GA-BPFD算法。该算法主要分两步:用遗传算法对测试报告进行预处理,得到误差最小的一组BP神经网络权值和偏置值;将所得权值和偏置值作为BP神经网络的初始权值和初始偏置值,用训练完成的神经网络结合测试报告进行系统级故障诊断。详述用于优化BP神经网络的遗传算法的具体步骤,对GA-BPFD算法的时间复杂度进行分析,并进行实验仿真。实验结果表明,相比BPFD算法,GA-BPFD算法具有较高的诊断精度、较低的时间复杂度和良好的泛化能力。

Genetic algorithm to optimize BP neural network diagnostic algorithm （GA-BPFD） was proposed for the system-level diagnosis based on Chwa ＆ Hakimi model. The algorithm consisted of two steps： the genetic algorithm was used to preprocess the test report to obtain a set of minimum error BP neural network weights and bias values; the weights and bias values obtained through previous step were taken as the initial BP neural network weights and bias values, then the faults were diagnosed according to the test report. The GA-BPFD algorithm was described in details, the time complexity was analyzed, and a simula- tion experiment was organized. The experimental results show that compared with BPFD algorithm, GA-BPFD algorithm has higher diagnostic accuracy, lower time complexity and good generalization ability.