Model-based automatic programming for plant control

A model-based automatic programming system for sequence control is described. The domain knowledge of the application area is discussed and an environment model that represents concepts of the domain-physical entities, actions and states, relations and conditions-in an extended semantic network is proposed. The system also has model-derivation methods. Because it uses a generic model to generate a specific description of the target plant, the system can be applied to the control of any plant of the same type. The model-based approach supports every task in the software life-cycle. As a result, it enables designers to specify control programs using natural terms and concepts. The system was evaluated by generating practical control programs for steel plants. It doubled design productivity while still producing programs that were satisfactory to design experts. These results suggest that at a primary goal of automatic programming research should be the development of models of the domain area.<
>