Abstract: Redundant data,which are useless to objective classification,are involved in calculation of membership degrees of the objective.Therefore,in order to classify objects properly,through mining the hidden knowledge information from the membership of index,a filtering algorithm,which can filter the redundant data in the index membership function,is proposed to converse membership degrees without no redundancy date.Based on this,at last,a general algorithm for conversing membership degrees which can be used in reliability evaluation on inertial navigation system simulation is proposed.