A hybrid artificial neural networks and knowledge-based expert systems approach to flexible manufacturing system scheduling

Flexible manufacturing system (FMS) scheduling is a complex problem in nature that leads to a high level of uncertainty due to limited feasible solutions in an extensive search space. Heuristics involving dispatching rules have been widely utilized to obtain good solutions. This strategy has been recently enhanced by FMS scheduling researchers using knowledge-based expert systems as means of resolving scheduling problems. Unfortunately, the knowledge-based expert systems (KBESs) developed are limited in real-time performance due to cracks in their encoded knowledge or lack of adequate plans to address the changing environment. A framework is developed displaying the capabilities of automatic learning and self-improvement, providing the necessary adaptive scheme to respond to the dynamic nature of flexible manufacturing systems. This proposed framework uses a hybrid architecture that integrates artificial neural networks and knowledge-based expert systems to generate solutions for the real time scheduling of flexible manufacturing systems. In this framework, the artificial neural networks perform pattern recognition and, due to their inherent characteristics, support the implementation of automated knowledge acquisition and refinement strategies through a feedback mechanism. They enable the system to recognize patterns in the tasks to be solved in order to select the best scheduling rule according to different criteria. The knowledge-based expert systems, on the other hand, drive the inference strategy and interpret the constraints and restrictions imposed by the upper levels of the control hierarchy of the flexible manufacturing system. The level of self-organization thus achieved provides a system architecture with a higher probability of success than traditional approaches.