Research On A Class Of Bi-level Programming Model And Genetic Algorithm For Product Family Design

Product family design is a complicated flow, it is not only the interactions in turnamong the customer domain, functional domain, physical domain and process domain,but also influenced by the supply chain design, products sale, after-sales service andso on. From the perspective of system theory, there are a lot of leader-followerrelationship between the stages of product family design or between product familydesign and its related system. Bi-level programming model is a kind of mathematicalmodel which reflects the leader-follower relationship in the economic andmanagement issues, it belongs to NP difficult problem, although there are manystudies of its algorithm, they often only focus on the bi-level programming withspecial properties. Genetic algorithm is one of more popular intelligent algorithms andpossesses the characteristics of simple generality, robustness, global convergence andsuitable for parallel processing, especially has a more widely application in the fieldof mathematical optimization. This dissertation puts forward the leader-follower jointoptimization problems that exist in product family design through introducing thegeneral process and influence factors of product family design, extracts a typical classof bi-level programming model from the existing study for product family design anddesigns a genetic algorithm method. At last, this dissertation studies a leader-followeroptimization problem in product family design and applies the model and solvingalgorithm in a specific example.The main work and innovations of this dissertation are as follows:Firstly, based on the general process of product family design, this dissertationputs forward a few kinds of leader-follower optimization problem of product familydesign, and extracts a value-type bi-level programming model for product familydesign.Then, based on the genetic algorithm, this dissertation designs a method solvingthe model. According to the solution concepts of bi-level programming, the algorithmadopts the nested genetic algorithms. The outer genetic algorithm solves the upperproblem. The inner genetic algorithm solves the lower level problem given the uppervariables. This algorithm can not only guarantee that the solution is within constrainsdomain, but also ensure that it approximates the optimal solution. Finally, based on the existing joint studies between product family moduleconfiguration and supply chain design, this dissertation applies the algorithm andmodel to their leader-follower optimization.