Bounding the computation time of forward-chaining rule-based systems

For real-time applications of expert systems, success depends on the computational efficiency of the implementation. In this study, we propose an analytical method for evaluating the processing time of forward-chaining rule-based systems. An upper bound based on this system model is developed. If the upper bound stays within the time available for planning the operational or control task, the expert system would be able to complete the rule-processing in time. To compute the upper bound, the worst case working memory element sets are obtained for each functional step of the matching procedure. The worst case time for rule selection in the conflict resolution step is also derived. The maximal number of firings for each rule is considered in order to arrive at a bound for total processing time. Numerical examples are presented which point out the importance of rule and data structures in the efficient implementation of rule-based systems.