基于RFECV-RF特征选择的W-SVM齿轮故障诊断与优化

Fault Diagnosis and Optimization of W-SVM Gear Based on RFECV-RF Feature Selection

为了实现对齿轮故障类型的准确诊断,提出了一种基于RFECV-RF特征选择的W-SVM故障诊断分类算法。为避免特征冗余带来的偏差,采用RFECV-RF算法对特征变量进行重要度评估,与多个特征选择方法进行了对比。基于数据所具有的不均衡性,对SVM分类开展算法优化,引入了加权支持向量机(W-SVM),使用网格搜索进行超参数优化,将改进的特征提取与W-SVM进行融合优化。实验表明,与传统的SVM分类器相比,所提出的优化算法的识别准确率自提升达到6.4%,改进效果显著。

Indoer to achieve accurate diagnosis of gear fault types,the paper proposes a W-SVM fault diagnosis classification algorithm based on RFECV-RF feature selection.To avoid the bias caused by feature redundancy,the RFECV-RF algorithm is used to evaluate the importance of feature variables,and multiple feature selection methods are used for comparison.Based on the unevenness of the data,the algorithm optimization of SVM classification is carried out by introducing Weighted Support Vector Machines(W-SVM)and using grid search for hyperparameter optimization.In addition,the improved feature extraction is fused and optimized with W-SVM.Experiments show that the proposed optimization algorithm achieves a significant improvement in recognition accuracy with a self-improvement of 6.4%compared to the traditional SVM classifier.