基于人工蜂群算法的支持向量机优化  

Support Vector Machine Optimization Based on Artificial Bee Colony Algorithm

作　　者：刘路[1] 王太勇[2] 

机构地区：[1]天津大学精密仪器与光电子工程学院,天津300072 [2]天津大学机械工程学院,天津300072

出　　处：《天津大学学报》2011年第9期803-809,共7页Journal of Tianjin University(Science and Technology)

基　　金：国家自然科学基金资助项目(50975193/E050302);国家高技术研究发展计划(863计划)资助项目(2007AA042005);国家科技重大专项资助项目(2009ZX04014-101-05)

摘　　要：支持向量机的分类性能在很大程度上取决于其相关参数的选择,针对该问题提出基于人工蜂群算法的支持向量机参数优选方法并将其应用于电机轴承的智能故障诊断.该方法采用分类错误率的倒数作为适应度函数,利用人工蜂群算法对支持向量机的惩罚因子与核函数参数进行优化.通过在多个标准数据集上的测试证明,与遗传算法等传统优化算法相比,人工蜂群算法优化的支持向量机能够克服局部最优解,获得更高的分类正确率,并在小数目分类问题上有效地降低运行时间.将该方法应用于实测轴承故障信号的识别,对轴承故障信号进行小波变换,提取各个频段的归一化能量作为特征向量,利用该方法对特征向量进行分类,同样获得较高的分类正确率.A classification performance of support vector machine is largely dependent on the choice of its parameters. A parameter optimization method based on artificial bee colony algorithm is proposed to solve this problem and applied to intelligent motor bearing fault diagnosis. In this method, the inverse of classification error rate is used as fitness value, and the artificial bee colony algorithm is used to optimize the penalty factor and kernel parameter of sup- port vector machine. Compared with genetic algorithm and other optimization algorithms on standard datasets, the proposed algorithm can overcome the local optimal solution problem and acquire higher classification precision, and it costs less running time on small classification number of classification problem. Then the proposed method is applied to the recognition of bearing fault signals. The wavelet transform is applied to the bearing fault signals and the normalized energy values of every frequency band are extracted to compose feature vectors. The proposed method is used as the classifier and high classification precision is acquired.

关 键 词：人工蜂群算法 支持向量机 参数优化 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]