Intelligent alarm method by fuzzy measure and its application to plant abnormality prediction

Traditional methods for diagnosis/prediction of plant abnormality rely on the acquisition of fragmentary knowledge or rules from experts and the application of the state vectors of plant to portions of the knowledge or rules thus obtained to derive a conclusion. Therefore as the complexity of the plant to be evaluated increases, the degree of uncertainty of knowledge obtained from experts increases and thus the accuracy of evaluation decreases. To solve this problem, this paper proposes an intelligent alarm method which uses a fuzzy integral model based on fuzzy measure that can provide an accurate model of human subjective evaluation mechanism. The intelligent alarm method is unique in that it can quantify the uncertainty of the information to be evaluated and of human being who evaluates. To achieve these objectives, this paper provides a simulation of the expert´s overall plant evaluation by using an “interpretation of plant state based on expert knowledge” and a “overall judgement of plant state which amplifies the essence of the state interpretation information”. The method was applied to the prediction of abnormality for a drying process of a chemical plant to demonstrate that abnormality prediction by a simulation method is possible, that even for a plant without abnormality the method can provide estimate of the potential of abnormality, and that the method is useful in alleviating the burden of the monitoring and checking work for plant