Abstract
Traditionally, control software for the discrete systems such as FMS and coal-handling system has been developed using general purpose language based on the heuristic algorithm obtained through experience. However, development and modification of the control software consume time and man-hours because of uncomprehensive experession. In order to resolve the problem, a new control method, named the Rule Based Control method, has emerged.
In this method, as the control logics are described by the IF-THEN rules as data, maintainability and comprehensiveness of control software are extremely improved. However, this new method has following three problems: (1) the suitable domain for the rule description is not clear, (2) procedure of the rule development for multi-objective and multi-constraint control problem is not established, (3) the quantitative metrics for the rule expressions do not exist. This paper concerns with these problems.
First we define the control problem of the discrete system as the problem selecting alternatives such as works and resources to satisfy many constraint conditions and objective conditions. For specification of this type of control problems, IF-THEN rules are suitable for expressing these conditions. Next, the Halstead's metrics, which is one of the metrics for the property of general purpose languages, is enhanced to deal with rule expression based on the mechanism of the inference system.
The proposed method described above is successfully applied to the control rules development of a coal-handling system. The effectiveness of the procedure is shown by comparing the rule solution with Pareto optimal solutions. Furthermore, the enhanced metrics are validated through the comparison between rule expression and Fortran program.
