Diagnosis expert system for the life time of transformers by using fuzzy Clustering

Abstract

1.Introduction In general, it is known that a life time of dry type transformer is reduced remarkably by partial discharges. The quality of insulation is kept constant as possible in the factory, but sometimes it is scattered by some reasons. The reliability of the products is much influenced by scatters of the quality. Therefore, it is very important to predict the life time of the products. So far it was done by measuring a withstand voltage. However, the withstand voltage test means a destruction of the products, so it is necessary to predict the life time by using undestructive tests. As an undestructive test, the magnitude of the partial discharge will be adaptable to an expert system for life time prediction. General expert systems were based on some threshold value, which did not have an objective meaning. So it is necessary to evaluate more objectively. By using an expert system with fuzzy clustering, it can be done to make a judgment taking the interactions of each data into considerat ion. Even if a characteristic of life time is not clear, this type of expert systems can establish an optimum expert with learning, and can watch not only the products but also the relationship between conditions of manufacturing machines and life. However, fuzzy clustering has a flaw that if data numbers increases, handling time becomes extremely long. So, it can not to do real time diagnosis of the products in manufacturing line. Therefore, we apply a high speed algorism using DSP (Digital Signal Processor) with high calculating speed to our expert system. In this paper, we suggest a diagnosis expert system for the life time of transformers by using fuzzy clustering and describe a method of prediction of life time in the manufacturing line.