Hits:

Indexed by:期刊论文

Date of Publication:2021-09-11

Journal:COMPUTERS & INDUSTRIAL ENGINEERING

Volume:128

Page Number:894-904

ISSN No.:0360-8352

Key Words:Distributed evolutionary algorithm; Flexible scheduling; Apache Spark; Large scale optimization

Abstract:As a typical combinational optimization problem, the scheduling problem widely exists in many real-world manufacturing industry applications. With the intensification of marketing competition, the increasing problem scale results in the huge exponentially solution space which leads to the unacceptable storage space and computation time delay. In this paper, we consider the large scale flexible scheduling problem and treat the expectation of makespan as the objective function. A distributed cooperative evolutionary algorithm (dcEA) applied on Apache Spark is proposed. First, the dcEA adopts dimension-based distributed model to decompose the population into several sub-populations lengthways and randomly. Second, the dcEA defines resilient distributed dataset (RDD) as sub-populations and performs the identical evolutionary optimization process for all RDDs. Then, the hdEA updates the global best solution by the improved cooperative co-evolution framework. As a typical and basic scheduling problem, 10 benchmarks and three super large scale instances of flexible job shop scheduling are adopted and tested to prove the superiority of proposed dcEA. The numerical results show that dcEA has better performance and lower computational complexity.