基于超球面支持向量机的丝杠故障诊断技术

Fault diagnosis technology for NC machine screw based on hyper-sphere support vector machines

针对高档数控机床丝杠故障样本不易获取以及样本分布不均的问题,提出了一种用小波包分解和超球面支持向量机进行分类的丝杠故障智能诊断技术。该方法将振动信号小波包分解后的频带能量作为特征向量,输入到超球面支持向量机分类器进行故障识别。通过改变相关参数,研究了模型参数选择在构造超球面支持向量机中的重要作用。试验结果表明,建立的超球面支持向量机模型能够有效地对机床丝杠故障进行诊断。

It was difficult to obtain fault samples and the samples were distributed unevenly in Numerical Control（NC） machine tool screw.To deal with these problems,a novel method for screw fault diagnosis based on wavelet packet decomposition and Hyper-Sphere Support Vector Machines（HSSVM） classifier was put forward.The decomposed frequency band energy of vibration signal was selected as feature vectors and was inputted to HSSVM classifier which realized faults pattern recognition.The important role of model parameters selection in HSSVM classifier constructions were studied by shifting correlation parameters.Test results showed that HSSVM classifier model structured could detect screw faults of NC machine tool effectively.