CAJ | 학술논문

针对在柴油机故障诊断中支持向量机核函数参数和惩罚因子的不同取值会影响到分类正确率的问题，提出利用和声搜索算法对支持向量机相关参数进行选择优化( HS-SVM )的方法。在该方法中将参数作为和声存于记忆库中，而将支持向量机分类正确率作为目标函数，则整个寻优过程即是寻找使函数值最大的和声所对应的解。通过对UCI中的2个数据集进行分类正确率测试，并与未优化的支持向量机和人工蜂群算法优化的支持向量机( ABC-SVM )测试结果对比，证明了该方法的优越性。最后，将该方法用于柴油机故障诊断，并将分类正确率与未优化SVM和ABC-SVM分类结果进行比较，进一步说明和声搜索算法优化的支持向量机( HS-SVM)既能获得较高的分类结果正确率，又能有效降低运行时间，即说明该方法具有一定的实用性。
침대재시유궤고장진단중지지향량궤핵함수삼수화징벌인자적불동취치회영향도분류정학솔적문제，제출이용화성수색산법대지지향량궤상관삼수진행선택우화( HS-SVM )적방법。재해방법중장삼수작위화성존우기억고중，이장지지향량궤분류정학솔작위목표함수，칙정개심우과정즉시심조사함수치최대적화성소대응적해。통과대UCI중적2개수거집진행분류정학솔측시，병여미우화적지지향량궤화인공봉군산법우화적지지향량궤( ABC-SVM )측시결과대비，증명료해방법적우월성。최후，장해방법용우시유궤고장진단，병장분류정학솔여미우화SVM화ABC-SVM분류결과진행비교，진일보설명화성수색산법우화적지지향량궤( HS-SVM)기능획득교고적분류결과정학솔，우능유효강저운행시간，즉설명해방법구유일정적실용성。
In view of the problem which different values of Support vector machine kernel function parame-ters and penalty factor will affect the classification accuracy in diesel engine fault diagnosis, this paper uses harmony search algorithm to optimize parameters of support vector machine( HS-SVM) . In the method, the parameters will be stored in memory as a backup, and support vector machine classification accuracy as the objective function, so the optimization process is to look for the solution of harmony with the biggest func-tion value. Through classification accuracy test of two UCI data sets and comparison with test results from unoptimized SVM and ABC-SVM, it demonstrates the superiority of the method. Finally, the method is for diesel engine fault diagnosis, and through comparison with unoptimized SVM and ABC-SVM, it further proves that HS-SVM can obtain higher classification accuracy, and reduce the running time effectively. Namely, it illustrates practicability of the proposed method.