A new learning algorithm of ARTMAP using the ART2 network is proposed in order to apply ARTMAP to fault diagnosis for plant systems. Conventional ARTMAP using the ART2 network can be inadequate to categorize a complicated evaluation data set. To solve the problem,we propose a new architecture of ARTMAP which has a vigilance parameter for each category in ART2. The proposed ARTMAP can solve the problem and improve the performance of categorization of a complicated evaluation data set.