Multilevel Decision-making Theories And Algorithms For Fuzzy And Multi-follower Situations

Driven by a new round of industrial revolution, new productive mode, industrial form, business model and economic growth point are generated in relation to manufacturing development. Since manufacturing enterprises are always distributed in a digital and networked manufacturing environmnet, multilevel decision-making techniques are applied to deal with decentralized management problems between decision entities distributed throughout a multiple level hierarchy with the aim of intelligent manufacturing decisions. Bi-level decision problems and tri-level decision problems are the typical cases of multi-level decision-making, which have motivated a number of researches on decision models, solution approaches and applications.Nowadays, multilevel decision problems become large-scale because of high-dimensional decision variables; also, related decision parameters and conditions always involve uncertain information that is imprecisely or ambiguously known to decision entities; moreover, there may exist multiple decision entities in each decision level and multiple decision entities at the same level may have a variety of relationships with one another. To support large-scale, uncertain and diversified multilevel decision-making, the thesis studies theories and algorithms for solving nonlinear and large-scale multilevel decision prolems, fuzzy nonlinear bi-level decision problems and tr-level decision problems with multiple decision entities in each level, then applies these techniques to handle a decentralized production planning and inventory control problem in supply chain management, and lastly develops an algorithm toolbox to support modeling and solving multilevel decision problems. Detailed reseach contents of the thesis are proposed as follows.(1) Present a particle swarm optimization algorithm for solving nonlinear and large-scale multilevel decision problems. Since traditional exact algorithmic approaches lack universality and efficiency, heuristics-based particle swarm optimization(PSO) algorithms have been used to generate an alternative for solving multilevel decision problems. However, existing PSO algorithms are limited to solving linear or small-scale bi-level decision problems, and cannot be employed to solve tri-level decision problems. To handle these issues, this study develops a novel PSO algorithm for solving nonlinear and large-scale bi-level programs, and then extends the PSO algorithm to solve tri-level decision problems. For the sake of exploring the algorithm performance, the proposed PSO algorithm is applied to solve 62 benchmark problems and 810 randomly constructed large-scale problems.(2) Present a fuzzy nonlinear bi-level decision-making algorithm based on interactive compromise-based rules. In modeling and solving a bi-level decision problem, parameters that are often imprecisely or ambiguously known to decision entities are often characterized by fuzzy numbers. However, existing solution approaches are limited to handling special fuzzy numbers, e.g. triangular fuzzy number, and solving fuzzy decision problems in the linear version or in a special situation where all decision entities share the same constraint conditions. To solve general fuzzy nonlinear bi-level decision problems, this study discusses related theoretical properties based on a commonly used fuzzy number ranking method, develops a compromise-based PSO algorithm and illustrates the effectiveness of the proposed PSO algorithm by numerical examples.(3) Give decision models and algorithms for tri-level decision-making with multiple followers. Two decision entities or more are often involved at the middle and bottom levels in a tri-level decision-making case; these entities are called multiple followers. Moreover, multiple followers at the same level may have a variety of relationships with one another. In this situation, the various relationships between multiple followers will generate diverse decision processes, which need to be described and solved by different decision models and algorithms; this is called tri-level multi-follower(TLMF) decision-making. To handle this category of deicison problem, this study first presents TLMF decision models due to diverse relationships between multiple followers and discusses related theoretical properties. It then develops a TLMF Kth-Best algorithm for solving TLMF decision problems; also, a fuzzy programming approach is used to evaluate the solution obtained. Lastly, a detailed case study on production planning illustrates the effectiveness of the proposed TLMF decision-making techniques in applications.(4) Solve a decentralized vendor-managed inventory decision problem in a three-echelon supply chain using multilevel decision-making techniques. This study applies multilevel decision techniques to handle a vendor-managed inventory(VMI) decision problem in a three-echelon supply chain comprised of one vendor, multiple distributors and multiple buyers, gives the optimal inventory level and replenishment frequency under VMI coordination, and discusses operations of the multilevel decision-making process involving interactive decision entities.(5) Develop an algorithm toolbox for modeling and solving multilevel decision problems. Based on aforementioned multilevel decision theories and algorithms, this study develops an algorithm toolbox for multilevel decision-making, which can be applied to assist decision entities in modeling and solving multilevel decision problems accurately and efficiently. This study uses the case study on production planning illustrates the effectiveness of the algorithm toolbox in modeling and solving multilevel decision problems.