基于GM(1,1)模型的齿轮故障识别预测研究

Research on gear fault recognition and prediction based on GM (1,1) model

针对齿轮故障诊断问题,提出一种基于灰色预测模型(GM(1,1))的齿轮故障识别预测方法。根据齿轮磨损情况将齿轮故障定义为4种状态:正常齿轮状态、齿根裂纹故障状态、齿面磨损故障状态和断齿故障状态。通过搭建齿轮故障诊断实验台,利用振动传感器采集齿轮工作过程中处于不同磨损状态时的振动加速度信号样本,利用小波降噪、小波分解重构等方法重构特征信号,得到去噪后的信号图像,进而提取振动加速度信号特征数值,并基于GM(1,1)建立齿轮故障识别预测模型。研究表明,该模型预测数值与真实值误差均值在0.56%~0.67%,相对误差均方根值在0.33%~0.43%,其中齿根裂纹状态时相对误差均值为0.67%,相对误差均方根为0.33%;断齿故障状态时相对误差均值为0.67%,相对误差均方根为0.39%;齿面磨损故障时相对误差均值为0.56%,相对误差均方根为0.43%。表明模型预测精度较高,可用于齿轮故障预测研究,为齿轮故障诊断提供了一种新的研究方法及理论依据。

Aiming at the problem of gear fault diagnosis,a method of gear fault recognition and prediction based on grey prediction model(GM(1,1))was proposed.According to gear wear conditions,gear fault was defined as four states:normal gear condition,tooth root crack fault,tooth surface wear fault and broken tooth fault.A gear fault diagnosis experiment platform was first constructed and vibration sensor acquisition was used to collect the vibration acceleration signal sample data of the gear in different wear states during the working process.Then by using the methods of wavelet de-noising and wavelet decomposition reconstruction,the characteristic signal decomposition and reconstruction of the five-layer wavelet packet were completed and the signal image after de-noising was obtained.After the vibration acceleration signal characteristic value was extracted,a gear fault identification model based on GM(1,1)prediction was finally established.The experimental results show that the mean error between the predicted values and the true values is between 0.56%and 0.67%,and the root mean square of the relative error is between 0.33%and 0.43%.The mean value of the relative error is 0.67%and the root mean square of the relative error is 0.33%when the tooth root is cracked.The mean value of the relative error is 0.67%and the root mean value of the relative error is 0.39%when the tooth is brocken.The mean value of relative error is 0.56%and the root mean square of relative error is 0.43%when the tooth surface is worn.With very high prediction accuracy,this model can be used in the research on gear fault prediction,providing a new research method and theoretical basis for gear fault diagnosis.