多PCA模型及SVM-DS融合决策的服务机器人故障诊断  

Fault Diagnosis of Service Robot Based on Multi-PCA Models and SVM-DS Fusion Decision

作　　者：袁宪锋[1] 宋沐民[1] 周风余[1] 陈竹敏[2] 

机构地区：[1]山东大学控制科学与工程学院,济南250061 [2]山东大学计算机科学与技术学院,济南250101

出　　处：《振动．测试与诊断》2015年第3期434-440,587,共7页Journal of Vibration,Measurement & Diagnosis

基　　金：国家自然科学基金资助项目(61375084);山东大学基本科研业务费资助项目(2014JC034)

摘　　要：针对轮式服务机器人驱动系统故障诊断问题,提出一种基于多主成分分析(principal component analysis,简称PCA)模型及支持向量机和DS证据理论(support vector machine and dempster-shafer,简称SVM-DS)融合决策的故障诊断方法,分别利用正常状态和故障状态下的传感器数据建立多个PCA模型。利用正常状态下的PCA模型实现故障的检测。传感器数据经多PCA模型特征提取后作为SVM的输入向量,实现故障的初步分离。基于混淆矩阵定义SVM的全局及局部可信度,并依据可信度值和故障初步分离结果完成基本概率分配函数的赋值,以实现SVM和DS证据理论在故障分离中的有效结合。实验结果表明,本研究方法能灵敏检测到机器人驱动系统故障的发生,故障分离平均正确率达92.6%,与传统单PCA模型的方法相比有更高的正确率和稳定性。To solve the fault diagnosis problem of the wheeled service robot driving system,a novel fault diagnosis method based on multi-principle component analysis(multi-PCA)models is proposed,which compounds with support vector machine(SVM)and Dempster-Shafer evidence theory(DS).Multiple PCA models are established using sensor data sampled in the normal and fault states,respectively.The normal state PCA model is used to accomplish fault detection.During the fault isolation process,the multi-PCA models are used to carry out feature extraction from sensor data,and the processed data are taken as the input vectors of SVM classifiers,which achieves preliminary fault isolation.The global and local confidence values of the SVM classifiers are defined based on the confusion matrixes.To realize the effective combination of SVM and DS,the basic probability assignment(BPA)is appointed by the integration of the confidence values and the preliminary fault isolation results.Experimental results indicate that sensitive faults in the robot driving system can be detected,with an average fault isolation accuracy of92.6%.Compared with the traditional single PCA model method,the proposed method has better performance in accuracy and stability.

关 键 词：服务机器人 故障诊断 主成分分析 支持向量机 DS证据理论 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TP242.6[自动化与计算机技术—控制科学与工程]