摘要
针对支持向量在分类过程中,特别是对于非线性可分问题,如果采用不同的核函数,支持向量机(SVM)可以构造不同的学习机器和分类模型,从而导致分类算法复杂且分类精度较低。研究了SVM的轴承诊断原理,测试诊断方案及原始测试数据的特征提取,最后进行了数据训练和测试实验仿真,然后对风力发电机组机械故障进行诊断,实验说明了改进后的SVM故障分类方法的可行性和有效性。在建立故障分类模型之后,采用网格搜索法、遗传算法、粒子群算法对支持向量机的惩罚参数C和径向基核函数参数γ进行优化选择,通过分析发现故障分类精度提升比较明显。
Using different kernel functions,SVM can construct different machine learning and classification models; so classification algorithm becomes complex and classification accuracy is relatively low. This paper studies the support vector machine( SVM) bearing diagnosis principle,feature extraction test and diagnosis scheme and the original test data,the data of training and testing of simulation experiments,as well as diagnosis of mechanical fault of wind turbine. Experiment shows feasibility through improved method of SVM fault classification and effectiveness. After establishment of fault classification model and usage of grid search method,through genetic algorithm and particle swarm algorithm of support vector machine parameters C and RBF kernel function parameter optimization,the analysis of fault classification accuracy is greatly improved.
出处
《兰州石化职业技术学院学报》
2017年第4期8-11,共4页
Journal of Lanzhou Petrochemical Polytechnic
基金
兰州石化职业技术学院科研基金项目(KJ2015-20)
