基于支持向量机分类算法的齿轮箱故障诊断
Gearbox Fault Diagnosis Based on Support Vector Machine Classification Algorithm

针对现有齿轮箱故障评价需要研究设备机理造成的效率底下，功能性不强的问题，提出了基于支持向量机分类算法的齿轮箱故障诊断方法。首先，对传感器收集到的振动信号数据进行分析，提取相关特征。然后，绘制不同传感器在不同状态下的振动信号时间序列函数，并对这些函数的特征进行了简要分析。其次，对数据提取了平均值，方差这两个用以描述振动数据的总体趋势的特征变量，以及峰度，偏度这两个对判断齿轮箱齿轮故障有着重要作用的特征变量，并利用MATLAB、SPSSPRO对每一组数据进行了特征数据计算。最后，利用孤立森林、朴素贝叶斯、支持向量机三种分类算法，分别对数据集进行模型求解，然后通过对比三个算法结果中的准确率、召回率和测试集、训练集之间拟合程度，得到支持向量机分类算法针对齿轮箱的故障检测最优。
A gearbox fault diagnosis method based on a support vector machine classification algorithm is proposed to address the issue of low efficiency and weak functionality caused by the need to study equipment mechanisms for existing gearbox fault evaluation. Firstly, the vibration signal data collected by the sensor is analyzed and relevant features are extracted. Then, the time series functions of vibration signals of different sensors in different states are drawn, and the characteristics of these functions are briefly analyzed. Secondly, two characteristic variables, mean and variance, are extracted from the data to describe the overall trend of vibration data, as well as kurtosis and skewness, which are important for determining gearbox gear faults. Feature data calculations are performed on each set of data using MATLAB and SPSS PRO. Finally, three classification algorithms, namely isolated forest, naive Bayes, and support vector machine, are used to solve the model for the dataset. Then, by comparing the accuracy, recall, and fit between the test and training sets of the three algorithms, the optimal fault detection performance of the support vector machine classification algorithm for the gearbox is obtained.