基于支持向量回归的多参数设备故障预测方法  

Method of Mechanical Equipment Fault Prognosis Based on Multi-parameter Support Vector Regression

作　　者：武兵[1] 林健[2] 熊晓燕[1] 

机构地区：[1]太原理工大学机械电子工程研究所,太原030024 [2]太原理工大学信息化管理与建设中心,太原030024

出　　处：《振动．测试与诊断》2012年第5期791-795,864,共5页Journal of Vibration,Measurement & Diagnosis

基　　金：国家自然科学基金资助项目(编号:50975188);山西省自然科学基金资助项目(编号:2012011046-10)

摘　　要：针对热连扎生产线的大型、复杂生产装备,构建了一个基于支持向量回归的无量纲参数智能故障预测系统。通过对已有的热连轧生产线在线监测信息进行无量纲统计分析,提出了利用支持向量回归技术对多个特征量进行并行预测,并对各类特征及其各预测步骤的剩余使用寿命预测结果进行趋势分析,通过置信区间估计确定最终预测结果,为管理者制定维修决策提供可靠依据。For the large, complex production equipment like the hot strip rolling production line, a fault prognostic system is proposed which is based on support vector regression and dimensionless multi-parameter. Most of the proposed methods of support vector machine prediction model are calculated based on the single-input characteristics. At first, it is processed by dimensionless statistical analysis of the online data acquisition on the hot rolling production line. The next a support vector regression is employed to parallel predict of each fault feature, and prediction results of each step are processed trend analysis of the remaining useful life. At last the final prognostic result is estimated by the confidence interval that can be provide a reliable basis for managers to develop more accurate maintenance decisions.

关 键 词：剩余使用寿命 支持向量回归 无量纲统计量 故障预测 

分 类 号：TH165.3[机械工程—机械制造及自动化]