Evolutionary multi-objective blocking lot-streaming flow shop scheduling with interval processing time

Graphical abstractThis illustrates the flowchart of our proposed algorithm. First, the population is partly initialized using solutions obtained by a variable single-objective heuristic search algorithm, and the rest solutions are randomly generated. Second, a new crossover operator is developed utilizing valuable information embedded in non-dominated solutions and differentiation between parents. Third, choose solutions at a certain probability from the temporary population, keep the better ones based on their distance to the ideal point to form the offspring population. Last, merge the temporary population with the parent populations into a combined population, perform non-dominated sorting and calculate the crowding distance. The better individuals are selected based on the Pareto dominance and the crowded distance to form the parent population of the next generation. In this way, the exploration capability of the crossover operator and the exploitation ability of the ideal-point assisted local search are both considered in this algorithm. Display Omitted HighlightsA conversion of the BLSFS problem with interval processing time is proposed.A variant of single-heuristic is incorporated in population initialization.A novel crossover operator is proposed.An ideal-point assisted local search is applied to improve the exploitation.It contributes to enhance the capacity of the algorithm in tackling uncertainties. A blocking lot-streaming flow shop scheduling problem with interval processing time has a wide range of applications in various industrial systems, however, not yet been well studied. In this paper, the problem is formulated as a multi-objective optimization problem, where each interval objective is converted into a real-valued one using a dynamically weighted sum of its midpoint and radius. A novel evolutionary multi-objective optimization algorithm is then proposed to solve the re-formulated multi-objective optimization problem, in which non-dominated solutions and differences among parents are taken advantage of when designing the crossover operator, and an ideal-point assisted local search strategy for multi-objective optimization is employed to improve the exploitation capability of the algorithm. To empirically evaluate the performance of the proposed algorithm, a series of comparative experiments are conducted on 24 scheduling instances. The experimental results show that the proposed algorithm outperforms the compared algorithms in convergence, and is more capable of tackling uncertainties.