Leader-follower Joint Optimization Problem And Study On Model Method For Product Family Design

Compared with single-level model or multi-stages method, some problems ofproduct family design are more suitable for using leader-follower model, for thecharacter of multi-agent, multi-level and multi-disciplinary. Aiming at theleader-follower joint optimization problem of product family design, theleader-follower model of product family is established. Then, the decompositioncoordination method about leader-follower model is studied. At last, two types ofproduct family leader-follower joint problem are presented. Simultaneously, everyresearch chapter goes with comprehensicve case study, including mathematicalcalculation example, universal motor case and bicycle case. The main works of thedissertation are as following:Firstly,2D analysis figure, the mathematics model and general classificationframework are dicussed.2D analysis figure is helpful for establishing bilevel model.The mathematics model introduces the synergy item into the leader-follower modelwith the product family background. The general framework gives a classification ofleader-follower model from the followers relationship among variables and strategysets.Sencondly, for the optimization of hierarchical, multilevel systems inengineering, analytical target cascading is an effective method. Aiming at a class ofbilevel programming problem whose lower-level problem is a concave programming,after replacing the lower-level problem with KKT condition, this dissertationproposes a decomposition method of analytical target cascading with bilevel structure,based on the analytical target cascading and the FDT table of objective andconstraints.Thirdly, based on the general formulation of product family model with oneleader and multi-followers, aiming at the coupled decision variables amongmulti-followers, a decomposition coordination method is given. Through addingdeviation item on the objective, the variables of multi-followers become independent.The relationship between the Pareto solution of original problem and the feasiblesolution of decomposition coordination problem is discussed. Fourthly, based on a typical product family design problem of moduleconfiguration and parameter scaling, a bilevel decision structure reveals the coupleddecision between them. A bilevel mixed0-1non-linear programming model isdeveloped and solved, comprising an upper-level optimization problem and alower-level optimization problem. The upper level seeks for an optimal configurationof modules and module attributes by maximizing the shared surplus of an entireproduct family. The lower level entails parametric optimization of attribute values foroptimal technical performance of each individual module.Fifthly, based on a typical product family design problem of module attributecombination, a leader-follower model is proposed between mandatory modules andoptional modules, where0-1variable is used for denoting the choice of attributelevel.The upper level seeks for maximizing the shared surplus and commonality indexof an entire product family. The lower level seeks for maximizing the utility ofoptional attributes.