Knowledge-based diagnostic system of turbine with faults using the blackboard model

The paper describes a diagnostic system based on a blackboard model for a steam turbine with multiple faults. The system has been built and tested with the exercise equipment of the turbine. The knowledge of diagnosis is divided into six separate models, i.e. knowledge sources, which may be rule based or procedural or neural network. Different rule based knowledge sources can utilize different inference engines. The detected data and the information for describing conditions of the turbine are evolved into a blackboard, which is organized as a hierarchy with three different layers. Each layer is used in the different task, and serves for corresponding knowledge sources. The diagnosis process of the turbine with faults has simulated the technique of data fusion (sensor fusion), which can yield global optimal diagnosis conclusions by local and concurrent computations. The merits of the diagnostic system are compared.