摘要
针对磨粒识别中的LS-SVM分类器性能参数难以选择的问题,提出一种改进的遗传算法(IGA)对其进行优化选择。该算法将轮盘赌选择法与最优保留法相结合,采用一种随世代数增加而不断自动调整的交叉概率和变异概率,既提高了收敛速度,又易得到全局最优解。基于IGA的LS-SVM分类器磨粒识别方法为,利用主成分分析法(PCA)优选磨粒特征参数,并将结果作为样本训练LS-SVM分类器;通过改进遗传算法优化分类器参数,并通过测试样本测试分类器性能。仿真实验结果表明,此分类器的分类精度高,分类速度快。
Aiming at the difficult to select performance parameters of Least Square-Support Vector Machine(LS-SVM) classifier in the wear particle pattern identification process, an improved genetic algorithm(IGA) was put forward to realize the optimization selecting of performance parameters. The roulette wheel selection and the best reserve selection were com- bined by this algorithm, and the autoregulatory crossover operator and mutation operator were adopted when the generation was increased, so the convergence speed was improved, and the global optimal parameters could be found rapidly. The wear particle pattern identification method of LS-SVM classifier based on improved genetic algorithm was that using the principal component analysis (PCA) to select with optimization the characteristic parameters of wear particles and using the results as sample to train the LS-SVM classifier. By improving parameters of genetic algorithm optimal classifier and testing samples, the performance of the classifier was tested. The simulated experimental results show that this LS-SVM classifier has exact and rapid classification capability.
出处
《润滑与密封》
CAS
CSCD
北大核心
2013年第1期14-18,共5页
Lubrication Engineering
基金
航空科学基金项目(20101996012)
关键词
改进遗传算法
最小二乘支持向量机
主成分分析法
磨粒识别
improved genetic algorithm
LS-SVM
principal component analysis
wear particle pattern identification