基于主成分分析及多维高斯贝叶斯的超声流量计故障智能诊断方法

提出了基于多维高斯贝叶斯模型的设备故障智能诊断流程,包括数据的筛选与结构化分析、数据的降维、模型的构建、诊断结果的检验与分析等。研究表明采用主成分分析方法进行降维时,不同的诊断对象在降维参数的选择方面存在较大差别,诊断效果因诊断对象和样本数量的不同而有所差异。利用公开发表的超声波流量计数据库对流程进行验证。结果显示:针对B型流量计进行280次、C型流量计进行550次智能故障诊断,故障状态的首选正确识别率分别达到99.3%和95.1%,较k-最近邻(KNN)聚类分析算法有一定的优势。

Application of Primary Component Analysis and Multivariate Gaussian Bayesian Method on Intelligent Failure Diagnosis of Ultrasonic Flowmeter

The intelligent failure diagnosis method for equipment based on multivariate Gaussian Bayesian model was proposed. The method included data screening and structural analysis,data dimensionality reduction, model construction, verification and diagnostic results analysis. When using principal component analysis method for dimensionality reduction, it was shown that the selection of dimensionality reduction parameters has great influence on diagnosis result. The diagnostic effect varied with the property and quantity of samples. A publicly published ultrasonic flowmeter database was used to verified the method. By performing 280 and 550 failure diagnoses on two type of ultrasonic flowmeters (type B and type C) respectively, it was found that the correct failure recognition rate were up to 99.3% and 95.1%. Compared with the nearest neighbor KNN clustering analysis algorithm, this failure diagnosis method shows certain advantages.