基于改进ABC算法优化的LSSVM多分类器组机械故障诊断模型  

Mechanical Fault Diagnosis Model Based on IABC Algorithm Optimized Multiple LSSVM Classifier Group

作　　者：李鑫滨[1] 陈云强[1] 张淑清[1] 

机构地区：[1]燕山大学工业计算机控制工程河北省重点实验室,秦皇岛066004

出　　处：《中国机械工程》2013年第16期2157-2164,共8页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(61172095;51075349);河北省自然科学基金资助项目(F2012203138)

摘　　要：为了提高复杂机械故障诊断的确诊率,提出了一种基于改进人工蜂群算法(improved artificial bee colony,IABC)优化LSSVM多分类器组的故障诊断模型。该模型利用多特征提取方法,获取了较为完备的时频域特征信息,同时选择具有较强搜索能力和快速收敛性的IABC算法优化了LSSVM分类器的参数,提高了分类效率,在诊断决策层,利用评估矩阵进行了多分类器诊断结果的融合决策。通过与传统方法的对比表明:该诊断模型不仅能获取完备的故障特征信息,而且能更快地获取LSSVM最优分类参数;同时,基于评估矩阵的融合决策能够充分考虑各子分类器的性能差异,保证了诊断决策的高效精确。多种数据仿真表明,该诊断模型适用于机械故障诊断。In order to improve the fault diagnosis accuracy,the paper presented a fault diagnosis model based on IABC optimized multiple LSSVM classifier group. It used multiple feature extraction methods to get the complete time--frequency domain features. Then IABC was utilized to optimize LSSVM parameters,because it had strong searching ability and fast convergence to improve classifica- tion efficiency. And in the diagnostic decisions stage, the criterion matrix was utilized to make the deci- sion of the multiple classifiers. Through the comparison with the traditional methods, the results show that it can obtain complete features information,and can get LSSVM parameters more quickly and ef- fectively. The fusion decision based on criterion matrix fully considering the differences of sub--classi- fiers performance, and it ensures the higher accuracy of diagnostic decisions. Finally, the various simu- lation results show that the algorithm can be used in mechanical fault diagnosis well.

关 键 词：时频域特征 改进人工蜂群算法 LSSVM多分类器组 评估矩阵 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TH133[自动化与计算机技术—控制科学与工程]