摘要
为提高模拟电路的软故障诊断能力,提出一种基于KPLS特征提取和WNN的集成诊断方法。首先利用KPLS良好的特征提取能力,构建故障样本集的主元特征集;然后,利用WNN解决复杂非线性问题的优势,建立主元特征集的故障识别模型;最后,由所建模型对各种故障模式进行诊断判定。Sallen-Key带通滤波器的仿真测试表明:该集成方法仅通过不到300次迭代计算即完成模型训练,诊断的总正确率达到96.7%,且9种模式中的6种达到100%正确率,从而验证了其可行性和有效性。
In order to improve the ability of soft fault diagnosis of analog circuits, an integrated diagnosis method based on KPLS feature extraction and WNN was proposed. First, the good feature extraction ability of KPLS was used to construct the principal element feature set of fault sample set; then, the advantages of WNN on solving the complicated nonlinearity problems was applied to establish the fault identification model based on principal element feature set; finally, each failure mode was diagnosed and determined by the built model. The simulation experiment of Sallen-Key bandpass filter shows that the integrated method just completes the training of the model by less than 300 iterations computation with the total correct rate 96.7%, and the correct rate of 6 modes in 9 modes reaches 100%, which verifies its feasibility and effectiveness.
出处
《中南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2014年第6期1841-1846,共6页
Journal of Central South University:Science and Technology
基金
国防预研基金资助项目(9140A17020307JB3201)
关键词
小波神经网络
核偏最小二乘
特征提取
模拟电路
故障诊断
wavelet neural network
kernel partial least square
feature extraction
analog circuit
fault diagnosis
