冷轧过程断带故障的诊断研究

Research on Fault Diagnosis of Belt Tearing in the Cold Rolling Process

在冷轧过程中,断带故障是冷轧工序的主要生产故障之一。针对冷轧过程断带故障的特点,提出一种基于核主元分析(KPCA)非线性特征提取和最小二乘支持向量机(LSSVM)分类的故障诊断方法。此方法采用KPCA理论将冷轧过程原始空间数据映射到高维空间,并在高维空间进行主元分析,从而降维、去相关性,得到冷轧过程非线性特征向量。将降维后的特征主元作为LSSVM输入进行训练和识别,根据LSSVM的输出结果判断冷轧过程工作状态与故障类型。仿真结果表明:基于KPCA非线性特征提取和LSSVM分类的故障诊断方法计算速度快,能有效地提取冷轧过程断带故障特征,识别断带故障类型。

Belt tearing fault is one of the main faults in the cold rolling process. A method of fault diagnosis based on nonlinear feature extraction with kernel principal component analysis （KPCA） and least squares support vector machine （LSSVM） is proposed. The normal dataset of the cold rolling process is mapped from its original space into a hyper -dimensional feature space in which employing the PCA to reduce dimension, and then the new uncorrelated nonlinear features are extracted which can be used as input to the LSSVM model. Condition of the cold rolling process can be moni- tored and types of belt tearing fault are identified based on output results of LSSVM. In the end, the presented method is applied to diagnose the belt tearing fault in the cold rolling process. The diagnosis results from the cold rolling process show that the method with low cost based on KPCA and LSSVM can not only extract belt tearing fault features effectively but also identify the fault type better.