油料装备故障检测数据融合研究  

The Research on Data Fusion for Fault Detection in Oil Equipment

作　　者：曾慧娥[1] 周庆忠[2] 胡为艳[3] 

机构地区：[1]重庆科技学院机械与动力工程学院 [2]后勤工程学院 [3]78088部队

出　　处：《重庆科技学院学报（自然科学版）》2014年第3期93-96,共4页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：国家自然科学基金项目(50206033)

摘　　要：为解决油料装备检测数据庞大、故障不易诊断的问题,对油料装备故障检测数据融合进行研究。提出基于神经网络的状态检测数据融合模型,采用三步训练法进行传感器验证。使用数据压缩技术,将整个数据集投射到低维空间,将模式识别和多元统计技术作为故障隔离的单个分类器,利用后验概率进行特定类Bayesian融合,执行融合中心与单个分类器的联合优化。提出基于阶乘隐Markov模型的动态多故障诊断方法,通过寻找最大后验配置实现多分类器动态融合。应用结果表明该方法可提高对油料装备故障的诊断率。In order to solve huge detection data and difficult fault diagnose in oil equipment, its fault detection data fusion is researched. The state detection data fusion model based on neural network is proposed using three -step training method for sensor validation. The data fusion process in fault diagnosis is discussed. The entire data set is projected onto a low- dimensional space using data reduction techniques. The classification techniques including pattern classification and multivariate statistical technique are used as individual classifiers for fault isolation. Class - specific Bayesian fusion is implemented using the posterior probability, and joint optimization of the fusion center and individual classifiers is complied. Dynamic multiple fault diagnosis process based on Factorial Hidden Markov model is proposed. Dynamic fusion of multiple classifiers is achieved via finding the maximum posteriori configuration. The application shows that this research plays an important role in improving the performance of fault diagnosis for oil equipment.

关 键 词：油料装备 故障 检测 数据融合 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]