A key characteristic parameter extraction method based on vector similarity

Abstract

The characteristic parameters of electrical equipment have great significance on the equipment fault diagnosis. The larger the amount of characteristic parameters is, the more accurate the result of fault diagnosis will be. However, with the number of characteristic parameters increasing, it will bring the problems of serious redundancy and large amount of computation. The characteristic parameters extraction methods could solve these problems. But different bases, which are used to extract the key characteristic parameters, have different importance for single equipment. General characteristic parameter extraction methods could not solve that problem. Thus, a key characteristic parameter extraction method based on vector similarity is proposed to solve the above problems in this paper. Firstly, need to determine the common characteristic parameters set and the bases for extracting key characteristic parameters and standard vector. These bases could be selected from the guidelines, regulations, standards, online monitoring parameters and so on. Construct the eigenvectors of the characteristic parameters, then calculate bases' weights and the similarity of every eigenvector and standard vector by using improved Jaccard coefficient. Utilize the F distribution to calculate the membership degree of each similarity. After determining extraction threshold, if it is less than one membership degree of similarity, the corresponding characteristic parameter would be selected as key characteristic parameter. But if on the contrary it would be removed. Finally, gradually extract the key characteristic parameters. Taking the converter valve as an example, this method is used to select 8 key characteristic parameters from 30 common characteristic parameters. In this paper, that extraction threshold is more appropriate between 0.65–0.85 is put forward.