作者：王印松, 吴军超

作者单位：华北电力大学控制与计算机工程学院, 河北 保定 071003

关键词：气动调节阀;故障诊断;主成分分析;贝叶斯分类;DAMADICS平台

摘要：

该文提出一种基于主成分分析（principal component analysis，PCA）和贝叶斯分类的故障诊断方法，并将其应用在气动调节阀的故障诊断中。首先，应用DAMADICS平台仿真气动调节阀多种易发生的故障，监测用于进行故障诊断的信号，采集诊断过程所需要的训练数据集和测试数据集，并对数据集进行主成分分析处理，降低其维度，进而获取数据集的主要特征；然后，利用极大似然估计方法求出训练数据集所满足的多元高斯分布的均值和方差，得到每种故障模式下训练数据集分布的概率密度函数；最后，应用测试数据集进行验证，对于测试数据集中的每个测试数据样本，分别计算测试数据样本属于各种故障类型的后验概率，后验概率越大，对应发生故障的可能性就越大。将该方法与支持向量机（support vector machine，SVM）诊断方法和k-近邻（k-nearest neighbor，k-NN）诊断方法进行对比，诊断准确度整体较高，方法可行。

A fault diagnosis method for pneumatic regulating valve based on PCA and Bayesian classification
WANG Yinsong, WU Junchao
School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, China
Abstract: In this paper, a fault diagnosis method based on principal component analysis and Bayesian classification is proposed and applied to the fault diagnosis of pneumatic regulating valve. Firstly, the DAMADICS platform is applied to simulate a variety of prone faults of pneumatic control valves, monitor the signals used for fault diagnosis, obtain the training data sets and test data sets required for the diagnosis process, and perform principal component analysis on the obtained data sets. Processing, reducing the dimensions of the data set, and then obtaining the main features of the data set. Then using the maximum likelihood estimation method to obtain the mean and variance of the multivariate Gaussian distribution satisfied by the training data set, and obtaining the the probability density function of the data set distribution under each failure mode. Finally, the test data set is applied for verification. For each test data sample in the test data set, the posterior probability of the test data samples belonging to various fault types is calculated respectively, and the posterior probability is larger, corresponding the greater the likelihood of failure. The method is compared with the support vector machine diagnostic method and the k-nearest neighbor diagnostic method. The overall diagnostic accuracy is high and the method is feasible.
Keywords: pneumatic control valve;fault diagnosis;principal component analysis;Bayesian classification;DAMADICS platform
2019, 45(12):112-118  收稿日期: 2019-06-06;收到修改稿日期: 2019-07-22
基金项目: 国家自然基金联合基金项目（U1709211）
作者简介: 王印松(1967-),男,河北河间市人,教授,博士,研究方向为先进控制策略和控制系统故障诊断技术
参考文献
[1] 谈斐祺.基于统计学习的气动调节阀故障诊断研究[D].杭州:浙江大学, 2016.
[2] PUIG V, STANCU A, ESCOBET T, et al. Passive robust fault detection using interval observers:Application to the DAMADICS benchmark problem[J]. Control Engineering Practice, 2006, 14(6):621-633
[3] ELAKKYIA V, KUMAR K R, GOMATHI V, et al. Investigation of fault detection techniques for an industrial pneumatic actuator using neural network:DAMADICS case study[J]. Advances in Intelligent Systems and Computing, 2015, 324:237-246
[4] HEYDARZADEH M, NOURANI M. A two-stage fault detection and isolation platform for industrial systems using residual evaluation[J]. IEEE Transactions on Instrumentation and Measurement, 2016, 65(10):1-9
[5] 赵铭, 金大权, 张艳, 等. 基于EM和GMM的朴素贝叶斯岩性识别[J]. 计算机系统应用, 2019, 28(6):38-44
[6] 王乐慈, 高世臣, 林孟雄, 等. 基于不同概率密度估计方法的朴素贝叶斯分类器[J]. 中国矿业, 2018, 27(11):174-180
[7] 何玉林. 基于核密度估计的光谱数据分类与回归方法研究[D].保定:河北大学, 2014.
[8] BARTYS M, PATTON R, SYFERT M, et al. Introduction to the DAMADICS actuator FDI benchmark study[J]. Control Engineering Practice, 2006, 14(6):577-596
[9] YIN S, STEVEN X D, NAIK A, et al. On PCA-based fault diagnosis techniques[C]//Control & Fault-tolerant Systems. 2010.
[10] 杜瑞杰. 贝叶斯分类器及其应用研究[D].上海:上海大学, 2012.
[11] 刘玉东. 基于SVM的气动调节阀故障诊断[D].杭州:浙江工业大学, 2016.
[12] GUO G, HUI W, BELL D, et al. KNN model-based approach in classification[J]. Lecture Notes in Computer Science, 2003, 2888:986-996