可变步长网格搜索法结合LIBSVM预测风机故障研究

Variable step-grid searching method combines the LIBSVM to identify fan fault

针对风机的智能故障诊断,采用了一种可变步长网格搜索法寻优结合LIBSVM工具箱实现风机故障智能识别的方法。LIBSVM的信号特征输入选用风机振动信号的小波能量谱,用交叉验证法验证识别准确率,用LIBSVM仿真建立支持向量机分类模型,最后用测试样本测试,发现基于该方法可以保证支持向量机识别率并且可以有效减少运算量。

Aiming at the intelligent fault diagnosis of fan,a method of variable step size grid search and LIBSVM toolbox is adopted to realize the intelligent fault identification of fan.The wavelet energy spectrum of fan vibration signal is selected as the signal feature input of LIBSVM,and the recognition accuracy is verified by cross validation method.The support vector machine classification model is established by LIBSVM simulation.Finally,the test sample is used to test,and it is found that the recognition rate of SVM can be guaranteed and the calculation amount can be effectively reduced based on this method.