Elite Archive-Assisted Adaptive Memetic Algorithm for a Realistic Hybrid Differentiation Flowshop Scheduling Problem

This research presents an original and efficient elite archive-assisted adaptive memetic algorithm (EAMA) to deal with a realistic hybrid differentiation flowshop scheduling problem (HDFSP) with the objective of total completion time minimization. In this scheduling problem, each job consists of multiple parts and the jobs are divided into different types. The manufacturing of a job is comprised of three consecutive stages: 1) parts fabrication on first-stage parallel machines; 2) parts assembly on second-stage single machine; and 3) job differentiation on one of third-stage dedicated machines. We provide a mixed-integer programming model, derive three lower bounds, and further present the EAMA metaheuristic for HDFSP. The EAMA is initialized heuristically, and its global exploration is performed by a differential evolution, which includes three newly designed operators: 1) elite-driven discretized differential mutation; 2) probability crossover; and 3) biased selection. To enhance the local search, an external elite archive is set and evolved in parallel with global exploration by a meta-Lamarckian learning-based adaptive multistage local search and a variable length-based adaptive block-insertion local search. After the global exploration and local exploitation, an elite sharing strategy is used to exchange the excellent information between population and elite archive, and an adaptive restart strategy is used to diversify the population. The influence of parameter setting on EAMA is surveyed by using an improved design-of-experiment. The statistical results from extensive computational experiments demonstrate the effectiveness of the special designs and show that EAMA performs more efficient than the existing algorithms in solving the problem under consideration.