模糊聚类在齿轮典型故障特征模态识别中的应用

Research on Diagnosis of Typical Gear Faults Based on Wavelet Power Spectrum and Fuzzy Clustering

运用MATLAB软件对齿轮箱振动加速度信号进行小波变换初步滤波,得到其尺度-功率谱,再利用模糊聚类的方法将齿轮故障信号的尺度-功率谱幅值向量进行分类。试验结果表明:该方法可应用于变速箱齿轮的故障诊断中,能够对齿轮3种不同运行状态(正常、磨损、断齿)进行准确分类,分类准确率达100%。

In this paper, the diagnosis of gear faults by fuzzy clustering based on wavelet transform was made to resolve the problem of extracting feature of gearbox vibration acceleration signal. In this process the wavelet transform was used to filter wave of the gear spectrum. The experiment result shows that three gear working states （normal, abrasion and tooth breaking） can be distinguished by this approach when it is used in the diagnosis of gear fault. The accuracy of classification reaches 100%.