Research On Intelligent Optimization Algorithms For The Distributed Permutation Flowshop Scheduling Problem With Preventive Maintenance

Due to the tight link with the multi-factory production environment in today’s decentralized economy,distributed scheduling problems have been a hot study topic.However,the existing literature rarely considers the effect of aging on production scheduling caused by machine running for a long time.This study investigates the distributed permutation flowshop scheduling problem,taking into account machine preventative maintenance.The main research work of this study is summarized as follows.First of all,a multi-start iterated greedy algorithm is proposed to minimize the makespan.By adding a dropout operation to NEH2,an improved heuristic for initialization and re-initialization is described,resulting in solutions with a high level of quality and dispersiveness.The simple but effective destruction and construction phases are applied for global search.The local search is integrated into the algorithm for neighborhood solution exploitation.In addition,a restart strategy is introduced to avoid local optima.Secondly,a hash map based memetic algorithm is presented to tackle the total flowtime minimization.The NEH based heuristic is incorporated with the DLR heuristic to provide the initial solution.The algorithm uses the hash map data structure to store all candidate solutions in the form of key-value pares.The selection,crossover,and mutation operators are also improved to broaden the scope of research in the discrete domain.The local search combines the job insertion and job swap operators to improve the obtained offspring solutions and makes use of the population hash map’s characteristic to shorten solution evaluation time.Thirdly,based on the previous research,the problem with the dynamic event of new job arrival is studied.A discrete artificial bee colony algorithm is presented to minimize both the makespan and total flowtime.The initial solutions are generated by an improved NEH heuristic with random operations.Different strategies are adopted in the three phases for population diversity and individual quality.The algorithm utilizes the fast-non-dominated-sort method for solution selection.Finally,the effectiveness of the proposed algorithm is verified by simulation and comparison with other advanced algorithms in different test cases.