Multi-criteria optimization of flexible job shop scheduling problem with hybrid metahevristic algorithm

Težišče proizvodnje se vse bolj nagiba od masovne proizvodnje k proizvodnji po željah kupca, torej k proizvodnji po naročilu. Načrtovanje in vodenje takšne proizvodnje je v konkurenčnih pogojih poslovanja izjemnega pomena. Kratki pretočni časi naročil, visoka zanesljivost dobavnih rokov, nizke zaloge, trajnostna naravnanost in ugoden vrednostno-časovni profil, vezan na tok vrednosti, postajajo ključni proizvodni cilji, ki jih lahko dosežemo le z ustreznim terminskim planiranjem. Stohastični prihodi naročil in različna zaporedja ter obseg obdelav pri naročilih lahko povzročijo zelo neenakomerno zasedenost kapacitet, kar privede do velikega raztrosa pretočnih časov operacij in odstopanj dobavnih rokov. Disertacija obravnava terminiranje proizvodnje po naročilu z uporabo hibridne metahevristične metode evolucijskega računanja tako, da rešuje obravnavan problem večkriterijske optimizacije z uporabo nove metode evolucijskega računanja, prilagodljivim simulacijskim modelom in ustrezno ovrednotenim vrednostno-časovno-prilagodljivostnim diagramom. Z analizo numeričnih in simulacijskih eksperimentov je dokazana visoka zmožnost reševanja optimizacijskih problemov, kjer optimizacijske rešitve predlaganega algoritma dokazujejo primerljivost z rešitvami dveh aktualno najuspešnejših algoritmov. Predstavljena je nova metoda prenosa optimizacijskih rezultatov v simulacijski model, ki omogoča popolno prilagodljivost in avtomatiziran vnos vhodnih parametrov, tako testnih kot realnih podatkov proizvodnih sistemov. Celovit optimizacijski pristop je ovrednoten z novo metodo vrednostno-časovno-prilagodljivostnega diagrama, ki omogoča podrobno analizo in potrditev ustreznosti rezultatov predlaganega optimizacijskega algoritma in simulacijskega modela. Kot dokaz zmožnosti reševanja realnih optimizacijskih problemov je v zaključnem poglavju predstavljena validacija celotnega pristopa na podatkih realnega proizvodnega sistema, kjer je potrjeno osnovno raziskovalno vprašanje pomembnosti večkriterijske optimizacije proizvodnje po naročilu.The focus of production systems optimization is moving increasingly from mass production to mass customization or make-to-order manufacturing. Production planning and scheduling of such production systems is very important, due to competitive business conditions. Short flow times of orders, high reliability of delivery dates, low stocks, sustainable manufacturing, high flexibility and a favourable cost-time profile, linked to the manufacturing value flow, are becoming the key production goals. They can be achieved mainly with appropriate multi-objective production optimization. Stochastic orders arrivals, different sequences and different production volume of orders, can lead to a very uneven capacity utilization, resulting in a longer flow time of operations and in the deviation of delivery dates. The dissertation deals with the flexible job shop production scheduling using the hybrid metaheuristic method of evolutionary computation for solving the problem of multi-objective optimization with a new method of evolutionary computation, adaptive simulation models and appropriately evaluated cost-time-flexibility diagrams. By analysing numerical and simulation experiments, it was demonstrated a high ability to solve optimization problems where the optimization solutions of the proposed algorithm prove the comparability with the solutions of the two currently most successful algorithms. A new method of transferring optimization results to a simulation model is presented, enabling full flexibility and automated input of production system parameters, used for both test and real production systems data. The comprehensive optimization approach is evaluated by a new method of cost-time-flexibility diagram, which allows detailed analysis and confirmation of the proposed optimization algorithm and simulation model results adequacy. As a proof of the ability to solve real optimization problems, the final chapter presents the validation of the whole approach on the real production system data, where the basic research question of the multi-objective flexible production system optimization importance has been confirmed