基于GARCH模型MSVM的轴承故障诊断方法

针对振动信号因非平稳性导致自回归(AR)模型无法有效描述信号特征的不足,提出一种基于广义自回归条件异方差(GARCH)模型多类支持向量机(MSVM)的故障诊断方法。该方法首先利用GARCH模型拟合各种故障信号,将所得模型参数作为故障诊断特征,以MSVM作为故障诊断方法。试验结果验证了GARCH模型方法的可行性和有效性,同时将该方法同基于AR模型的方法及其改进方法进行比较,结果表明该方法在诊断率及诊断时间上都有明显提高。

Bearing fault diagnosis with a MSVM based on a garch model

In bearings fault detection application, to solve the problems of difficultly describing signal characteristics using auto-regressive(AR) models due to the signal’s non-stationary time series, a novel one-class diagnosis model using Multi-class Support Vector Machines based on Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model is presented in this paper. The GARCH model is employed to describe the different bearings vibration signals, the parameters of the generated different models is used as the signal’ diagnosis characteristics and consequently the classifier based on MSVM is established. The experiments show that the proposed approach can efficiently overcome the drawback of conventional methods based on AR’ parameters. The proposed diagnosis scheme is compared against the conventional methods based on the AR model’s parameters and the other modified methods based on AR model as the diagnosis feature in the experiments. The results illustrate effectiveness of the investigated techniques with some concluding remarks in terms of diagnosis times and diagnosis rate.