Simultaneous Balancing And Sequencing For Mix-model Assembly Line With Multiple Objectives Via An Estimation Of Distribution Algorithm

Due to the growing competition and diverse requirements,the lean and flexible production mode are desired by each enterprise and therefore the mass production of one product model are replaced by the customized production of multiple products within assembly lines.In mixed-model assembly lines,the workload balance of different stations and model sequence of multiple products,are the two main problems related to the performance of the assembly lines.They need to be considered simultaneously when new products are developed or customer demands changes.Hence,this paper focuses on the following aspects:First,the characteristics of the line and its workstations are analyzed.Furthermore,the general mathematical model of concurrently balancing and sequencing for the mixed-model assembly line is established via a mixed integer linear programming approach.Second,various evaluation criteria are discussed and three indicators are selected: the absolute load deviation among different workstations,cycle time and total over-bound distance,considering the sequence-dependent idle times in all stations.Moreover,a multi-objective optimization approach is proposed based on the Pareto concept and a multi-objective decision-making method,to carry out integrated search for global solutions and provide parallel search for partial solutions.Next,an improved estimation of distribution algorithm(m-EDA)is designed for simultaneously solving the proposed balancing and sequencing problem.Specifically,the workstation-based encoding for balancing and permutation-based encoding for sequencing are utilized to enhance the searching performance;multiple heuristic rules are adopted to increase the representativeness of the initial population,and consequently the m-EDA can sample high quality and diversified individuals from the probability model;greedy local search operators are also employed to current individuals for better potential searching space;at last,the diversity maintaining mechanism is introduced in the process of individual selection and population updating so as to escape the local optimum and achieve the global optimal or near optimal eventually.In the end,in terms of four evaluation criteria including non-dominating rate,generation distance,the number of Pareto optimal solutions,and the uniformity of distribution,the performance of proposed m-EDA is compared with the classic algorithm NSGA-? and multi-objective ABC algorithm.According to experimental results obtained from independent running on a series of benchmark examples,the proposed m-EDA outperforms the two compared algorithms with respect to convergence and diversity.