基于优化SVM-DT的阀门故障诊断方法

Method for valve fault diagnosis based on optimized SVM-DT

针对过程控制系统中阀门故障种类多、类型相似、故障数据有很强的非线性等问题,提出一种基于支持向量机决策树(SVM-DT)的阀门故障分类算法,利用改进的遗传算法对识别率影响较大的支持向量机(SVM)参数C和σ进行优化。根据类间分离测度,找出分离测度值最小的两类样本,用优化后的支持向量机进行训练,将其合并为一个类簇与其它故障种类一起,找出最难分类的两类故障样本集,逐步将所有训练样本生成决策二叉树模型。通过动态执行器基准平台(DAMADICS)阀门模型对支持向量机决策树模型进行验证,实验结果表明,该算法能够有效提高对阀门故障的识别率,有效解决了传统的一对一(1-a-1)、一对多(1-a-r)算法中存在部分样本数据无法识别的缺陷,对工业控制系统中阀门的故障诊断有指导意义。

According to diverse and similar valve faults,and nonlinear fault data in process control system,an optimized support vector machine and decision tree （SVM-DT）algorithm for the valve fault classification was proposed.The improved genetic algo-rithm was used to optimize support vector machine （SVM）parameters C and sigma that had a greater influence on recognition. According to the separation measure between different categories,the optimized SVM was used to train two kinds of samples with minimized separation measure value,and they were merged into a cluster together with other faults.The most difficult clas-sification of two kinds of fault sample set was found out,and binary tree model was generated through all the training samples gradually.The SVM-DT algorithm was verified by the dynamic actuator reference platform （DAMADICS）valve model.Experi-mental results show that the proposed algorithm can improve the recognition rate of valve fault effectively,and the defects of tra-ditional one-against-one and one-against-all algorithm that part of sample data cannot be identified were solved.The proposed al-gorithm has important guiding significance for valves fault diagnosis in the industrial control system.