BOILER TUBE LEAKAGE DETECTION EXPERT-SYSTEM

Efficient and reliable operation is the main requirement of the modern power plant: The most probable reason for failure in the power plant boiler is tube leakage. It is usually detected when urgent action is n eeded to prevent accidents in the plant. Advance detection of boiler l eakage is of primary interest to secure maintenance planning and preve nt the adverse effect of tube rupture. The development of the tube fai lure detection system is a demanding issue for the large power plant b oilers. The present paper describes the development of an expert syste m for detecting boiler tube leakage. The system is based on selected d iagnostic variables obtained by radiation heat flux measurements. A se nsitivity analysis of the diagnostic variables is performed. A three-d imensional mathematical model of the boiler furnace is used to obtain the confidence level for the minimum leakage to be detected. The desig n of the expert system is based on relative values of the radiation fl ux reading as the diagnostic parameter. The leakage detection expert s ystem is designed in the knowledge base environment, comprising the kn owledge base containing facts, information on how to reason with these facts and inference mechanisms able to convert information from the k nowledge base into user requested information. The knowledge base is b ased on the object-oriented structure with the definition of the objec t LEAKAGE. The object class LEAKAGE is composed of subclasses SENSOR a nd CASES. The inference procedure uses a set of procedural processes i n the preparation of diagnostic variables reading for the decision mak ing process. A fuzzification process is used for conversion of actual diagnostic values into semantic values. The several steps of the infer ence procedure lead to the logic processing of individual and collecti ve representation of diagnostic variables represented in the knowledge base. A number of examples are given for leakage detection based on t he expert system reasoning and monitoring representative situations wh ich are imminent to the set of parameters describing situations preced ing boiler tube rupture. (C) 1998 Elsevier Science Ltd. All rights res erved.