基于主成分分析的液压泥炮系统泄漏监测  

Application of principal component analysis to monitoring of oil leakage in a hydraulic clay gun system

作　　者：邹俊[1] 傅新[1] 黄硕[1] 张哲谦[1] 杨华勇[1] 任玉军[2] 

机构地区：[1]浙江大学流体传动及控制国家重点实验室,浙江杭州310027 [2]上海宝钢集团公司,上海201901

出　　处：《浙江大学学报（工学版）》2009年第2期197-201,共5页Journal of Zhejiang University：Engineering Science

基　　金：国家“973”重点基础研究发展规划资助项目(2006CB705405);国家自然科学基金资助项目(50705082)

摘　　要：针对液压泥炮系统泄漏量小、故障频发的特点,将主成分分析(PCA)方法引入到液压泥炮系统的泄漏检测中.与传统的故障检测方法相比,PCA方法采用HotellingT2和Q统计作为故障检测的依据,具有不依赖过程数学模型的特点.基于多维数据驱动的PCA方法,建立了液压泄漏监测装置,利用核密度法生成T2和Q两个统计量的联合置信空间,提高了系统泄漏故障诊断的正确率.将该装置应用到宝钢三号高炉泥炮系统中,结果表明,泄漏监测系统能够及时、准确地发现系统泄漏故障.Aiming at the characteristics of little leakage and frequent failure in a hydraulic clay gun system, principal component analysis （PCA） method was introduced for the fault detection of hydraulic leakage. Compared with the model-based approach, the PCA-based method utilizes Hotelling T2 and Q statistics as rules for fault diagnosis, thus it is simple and straight-forward. Based on the data-driven PCA method, an oil leakage monitoring device was developed. In order to enhance the accuracy rate of diagnosis, the Kernel density estimation was adopted to establish the joint degree of confidence. Application of the hydraulic clay gun system in Baosteel blast furnace No. 3 verified the good performance of the leakage monitoring device. The oil leakage fault could be detected timely and rightly.

关 键 词：主成分分析(PCA) 核密度估计(KDE) 故障监测 液压泥炮 

分 类 号：TH137[机械工程—机械制造及自动化]