A fuzzy logic expert system for detecting generator H2 leaks

We describe the results of a cooperation between the Siemens Power Generation Group and Siemens R&D to develop a program for detecting H₂ leaks from the generator. This program is part of the comprehensive DIGEST (Diagnostic Expert System for Turbomachinery) analysis and diagnostics system for steam turbine generators in power plants. This modular system comprises algorithmically oriented design and analysis procedures with the knowledge based diagnostics modules based chiefly on the experience accumulated by experts over many years. The latter is characterized by an initially incomplete knowledge base which is then expanded and stabilized by the experience gained in operation. A further problem for this system is that this knowledge is not formalized, but rather is available only as rules formulated in natural language. This manner of expression, which is adequate for experts, implies the imprecision and indistinctness inherent in natural language. The complexity of the problem, the necessity for developing a knowledge base step-by-step and the requirement to express imprecise knowledge all lead to the decision to implement an expert system based on fuzzy logic