应用AR模型的多参数与多测点信息融合的故障分类
Fault Classification based on Multi-parameter and Multi-point Information Fusion of AR Model

为了找到针对齿轮传动系统多类故障分类的有效方法，对行星齿轮传动系统进行故障实验，获取振动信号。采用EMD方法对该振动信号进行预处理，得到若干个IMF分量之和，对前4个有效的IMF分量分别建立AR模型，得到对应的自回归参数序列φ，进而对其分别计算关联维数、最大Lyapunov指数、样本熵这3个混沌特征参数，并将其作为辨识特征量。将不同测点对应的φ的不同混沌特征参数信息融合作为支持向量机的输入向量，建立6种不同故障状态的训练集，实现对故障类型进行分类。结果表明：对实验获取的振动信号进行EMD和AR模型处理后，能在很大程度上提高故障分类准确率。
In order to find an effective method of fault classification of gear transmission system, fault tests are conducted on the planetary gear transmission system to acquire vibration signals. Empirical mode decomposition (EMD) method is used to process the vibration signals. The sum of several intrinsic mode function (IMF) components are obtained and the auto regressive (AR) model of the former four IMF components is established. and the regression parameter sequence is obtained, then the correlation dimension, maximum Lyapunov exponent and sample entropy are calculated, and these three chaotic characteristic parameters are used as fault identification features. The information of different chaotic characteristic parameters of different measuring points are fused and fed as input vector of support vector machine (SVM) to establish six kinds of different state of the training sets, then the classification of fault type can be achieved. The results indicate that the use the experimental vibration signals proceeded by EMD and AR modeling, the fault classification accuracy can be improve