基于遗传算法与最大最小原理的故障模式特征选择.PDFVIP

　　　　　　　 20 2　　JOU RNAL OF NAT IONA L UN IVERSIT Y OF DEFENSE TECHNO LOGY　V o.l 20N o. 2 1998 谢涛　张育林 (　　410073) 　　　　, 。, ; , 。 , , , 。, 。 　、 、 、 、 、 　V 433 - - MaxMinPrincipleBased Selection for theOptimalFeatureParameters in FaultDiagnosisUsingGeneticAlgorithms X ie Tao　Zhang Yulin ( , , , 410073) Departm ent ofA erospace Technology NUDT Changsha 　 Abstract M uch inportance has been attachad to the selection of optim al feature param eters sub- set in the fault diagnosis fields such as liquid rocket propulsion system. This paper presents an effec- tive m ethod of selection for the optim al feature param eters subset using GeneticA lgorithm s and based , on them ax im um andm inim um clustering criterion for sam ples so that the selected feature param eters subset can be used to com pose a sim plified real-tim e fault classifier w ith high robustness to various sorts ofnoises and disturbances. First, a com posite directionaldivergence index for sam ples is proposed ; , as an evaluation criterion for the selected feature param eters subset for fault diagnosispurpose then GeneticA lgorithm has been m odified in parts for this specifi cpermutation problem, the dynam i cfit- ness adaptation technique and all-sharing function are introduced in order to avoid the population s prem ature convergence. A n ad-ho cgeneti coperator is specially designed to im prove the feature selec- . , tio