A blackboard-based expert system for engineering design of electrical transformers

We have constructed an expert system to facilitate the engineering design of electrical transformers. From a user´s specification, the system should produce a fabrication description of the targeted product which conforms with specifications and constraints, and which offers a minimal fabrication cost. Different versions of our system were developed, but the last one is on a generic system for blackboard application, ATOME from Centre de Recherche en Informatique de Nancy (France). The application is based on blackboard architecture, a set of sources communicate through a common data structure, the blackboard. The knowledge sources are structured in a three level hierarchy: strategy, tasks and specialists. The control of the problem solving is distributed among the first two levels which are operating in a rule-driven mode. The third level concerns the specialists which contain domain knowledge and are directed by their respective task and are accessing directly the blackboard for creating, updating or deleting hypotheses (facts). The design process is executed in a forward chaining mode by the ATOME inference engine. In every cycle, the engine determines, through a hybrid multistage algorithm, the actions to be initiated. Ultimately the process will stop when no more rules are available at the strategy level; at this point, the blackboard contains the current solution. After presenting the different components of this design application, and in this context, we discuss the advantages of the blackboard model at the different levels of knowledge base structure and reasoning process