少量样本下基于PCA-BNs的多故障诊断  

Multiple fault diagnosis based on PCA-BNs with a small number of samples

作　　者：王进花[1,2,3] 马雪花 岳亮辉 安永胜 曹洁 WANG Jinhua;MA Xuehua;YUE Lianghui;AN Yongsheng;CAO Jie(College of Electrical&Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou University of Technology,Lanzhou 730050,China;National Experimental Teaching Center of Electrical and Control Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Gansu Huacheng Construction and Installation Engineering Group Co.,Ltd.,Lanzhou 730070,China)

机构地区：[1]兰州理工大学电气工程与信息工程学院,兰州730050 [2]甘肃工业过程先进控制重点实验室,兰州730050 [3]兰州理工大学电气与控制工程国家实验教学中心,兰州730050 [4]甘肃华成建筑安装工程有限责任公司,兰州730070

出　　处：《振动与冲击》2024年第4期288-296,共9页Journal of Vibration and Shock

基　　金：国家自然科学基金(61763028,62063020);甘肃省自然科学基金项目(20JR5RA463)。

摘　　要：针对一些工业设备因有标签故障样本数据少而导致诊断准确率低的问题,提出了一种PCA-BNs主成分分析和斯网络(principal component analysis-Bayesian networks, PCA-BNs)结合的多故障网络模型的建模方法。通过PCA对时序信号进行降维,得到相互独立的故障特征,提高提取故障关键信息的能力;利用融合单故障贝叶斯网络构建多故障贝叶斯网络结构的方法,解决BN建模过程耗时的问题;通过高斯分布与极大似然估计结合的方法确定网络参数,提高少量数据BN建模的精度,实现在少量样本下的故障诊断。试验结果表明,基于PCA-BNs的故障诊断方法在少量样本条件下,能实现高精度的故障诊断,并且有效缩减了算法运行时间。Aiming at the low diagnosis accuracy of some industrial equipment due to lack of labeled fault sample data,a modeling method of multi fault network model based on PCA-BNs principal component analysis and Bayesian network(PCA-BNs)was proposed.The dimensionality of time series signals was reduced by PCA to obtain independent fault features and improve the ability of extracting key fault information.The method of fusing single fault Bayesian network to construct multi fault Bayesian network structure was used to solve the time-consuming problem of BN modeling process.The combination of Gaussian distribution and maximum likelihood estimation was used to determine the network parameters,improve the accuracy of BN modeling with a small amount of data,and realize fault diagnosis under small samples.Experimental results show that the PCA-BNs fault diagnosis method proposed in this paper can achieve high-precision fault diagnosis under the condition of small samples,and effectively reduce the running time of the algorithm.

关 键 词：工业设备 故障诊断 时序信号 贝叶斯网络 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TH133.33[自动化与计算机技术—控制科学与工程]