Vyuziti prvku umele inteligence pri rozvrhovani vsadkovych vyrob

This dissertation is dedicated to the problems of optimal production scheduling in the batch chemical plants. The goal of scheduling is to assign equipment units to the tasks in an unambiguous way as well as to determine the exact timing of operations so as to optimize given objective criterion. The scheduling problems demand a large amount of time for computation and together with many other combinatorial problems belong to the NP-complete class from the mathematical point of view. Computational time increases exponentially as the dimension of the input data increases. Therefore solving algorithms must be rated on the basis of their ability to find out optimal schedule as well as to provide reasonable results in realistic times. Application possibilities of the three modern non-deterministic methods -tabu search heuristic, genetic algorithms and simulated annealing - which represent an alternative to the standard integer programming techniques (MILP) are analyzed in this dissertation work. Rules for effective adaptation of each method to the multiproduct scheduling problem are derived from the results of detailed sensitivity analysis of key parameters. New modifications of these methods as well as a new solving strategy for the short-term scheduling problem arising in multipurpose batch plants are proposed and verifiedAvailable from STL Prague, CZ / NTK - National Technical LibrarySIGLECZCzech Republi