Track: Production Planning and Management

Abstract

Abstract. A Dynamic Flexible Job Shop scheduling Problem (DFJSP) with different priority levels of resources is one of the most complex problems of machine scheduling. This problem is one of NP-Hard problems for which solving numerous heuristic and metaheuristic methods have so far been presented. Genetic Algorithms (GAs) are an effective method that gives quality solutions in reasonable computational time. In these approaches, improvement quality of solutions, avoiding premature convergence, and robustness of solutions is still among the challenging arguments and in the researchers’ center of attention. Hybridization is an effective way of improving the performance of GAs. Local search techniques are the most common form of hybridization that can be used to enhance the performance of GAs. Also adapting of GA operators (crossover and mutation) in amount and range of coverage and in the form of intelligent agents can operate as an efficient approach in improving its effectiveness. In this approach, the adapting in amount of operators prevents its premature convergence and the adapting in the coverage range of operators causes maximum use of the problem’s important resources. One of these important resources is Bottleneck Resource (BR). In the proposed GA (LSGAIA), initially the adapting in amount of operators based on the solutions’ tangent rate is done. Subsequently the adapting in the scope of operators, in first step, happens by converging operators on BRs (which was detected initially) and, in second step, occurs by converging operators on elite solutions. The resultant of all these heuristics causes that the search process (by having diversity) focuses on more probable areas than feasible region. The results in the static state are compared with other well-known algorithms used for the purpose from open literature. These comparing results indicate that the proposed LSGAIA is quite effective in both aspects of solution quality and algorithm robustness.