Simultaneous Balancing And Sequencing Of Mixed-model Assembly Lines Via Rule Extraction And Local Search

Mixed-model assembly lines,which are widely used in manufacturing enterprises,can produce multi-variety and small-batch products.However,most of the researches focus on the balancing problem or sequencing problem within mixed-model assembly lines,and relax the coupling relationship between balancing and sequencing problems,which results in the loss of the optimal balance and sequence scheming.Since the balancing and sequencing of mixed-model assembly line are strong NP-hard problem in combinatorial optimization,it is of great theoretical research value to study the effective algorithms to tackle this problem.Therefore,this paper attempts the following work:Firstly,the mixed-model assembly balancing and sequencing problem is analyzed,and then mixed integer linear programming method is used to model the problem according to the various constraints in actual production.Secondly,for the mixed-model assembly balancing and sequencing problem,gene expression programming algorithm is first applied to extraction rules,This paper analyzes the characteristics of the balance problem,and refines five heuristic factors,such as the candidate operation time,the direct follow-up operation,all subsequent operations,the demand of the ordered model product and the work load of the waiting model.and the current Heuristic factors with problem characteristics are introduced.In the process of rule extraction,the expression of gene expression encoding,sequence traversal tree decoding,balancing and sequencing scheme,fitness evaluation and evolutionary operation is developed,which effectively improve the performance of the algorithm.Thirdly,an improved multi-objective simulated annealing algorithm with rules is proposed to optimize the balancing and sequencing problem.In the initialization,Multiple heuristic rules are used to generate balanced balance and sequence schemes to improve the representation of initial solution sets.In the neighborhood search,two kinds of local search operators are proposed to optimize the current solution constantly.In the acceptance strategy,a diversity selection mechanism and a restart mechanism are added to avoid the algorithm trapping into local optimum and eventually reach a global optimum.Finally,the benchmark instances of mixed-model assembly line is solved and analyzed.in order to test the performance of different heuristics,the comparison experiments for different heuristic rule are designed.The statistical analysis tool ANOVA is employed to analyze the performance of the experimental results.In addition,four evaluation metrics,including generation convergence,spacing metric,the ratio of non-dominated solutions,and maximum spread,are utilized to evaluate the compared results with the classic algorithm NSGA-II and the multi-objective ABC algorithm.And these results based on a series of benchmark instances indicate that the MOSA algorithm is superior to these two comparison algorithms in terms of convergence,distribution and ductility.Moreover,to promote the practical application of the algorithm and develop the related system,the user interaction interface and system rule extraction are designed.