Reactive scheduling methodology in industrial application

Practical scheduling usually has to react to many unpredictable events anduncertainties in the production environment. Reactive scheduling incontrast to predictive scheduling can be dealt with such unpredictableevents and find a good solution for it by revising the initial schedule, butpredictive scheduling has a fixed schedule over the time that can not react to such events.In this thesis, case-based reasoning is applied to reactive scheduling tomodel the experience of skilled human operators in reacting to such eventsin the production environment. Case-based reasoning is supported byfuzzy reasoning on a constraint-based representation of schedule to findand adapt old cases and to evaluate them.We take the initial schedule from the real plant, then we used fuzzy settheory to represent its constraints and to find an appropriate function torepresent it and to evaluate its satisfaction value compared by a threshold value which is obtained from previous related work in the steel-making.The evaluation function is used to measure the degree of satisfaction of the whole schedule by weighting different constraints, this evaluation function is used also to measure the ability of CBR system to find a solution to unexpected events and its levels reaches to (0.993) .We finally compare the results of CBR with Tabu search as an iterativemethod, finding that CBR is more response than Tabu search, but Tabusearch may be given best value of evaluation function, so that we confirmthat the both method will be the best method to have optimal value of theschedule. We demonstrate the proposed methodology by introducingscheduling problem of steel-making plant in a real plant (HADISOLB)company.