The premise of Apriori algorithm is that the frequency and significance of each item in the database are equal or similar, but in the actual practice of fault diagnosis it's not the case. This paper improved it on the basis of Apriori algorithm, utilizing multiple minimum supports to solve the mining of non-frequent item in equipment fault diagnosis; meanwhile a CCWMMS Algorithm based on “credit component values” is proposed, which focus on the item's inconsistent degree of significance in the actual application. This paper also proved this algorithm's exactness and validity in fault diagnosis by actual examples.