Multi-objective Optimization Method For Distributed Precast Component Flow Shop

To accelerate the development of green and low-carbon,the country has vigorously promoted the application of prefabricated buildings,and thoroughly implemented the "Fourteenth Five Year" Construction Industry Development Plan and other relevant documents.As the main link of the prefabricated building industry,the precast component production is transforming to a distributed structure based on multi workshop and decentralized,which makes the distributed production mode of great research significance.Furthermore,production scheduling is an NP-hard problem in the field of optimization,and its complexity is reflected in large scale,multiple constraints,multiple objectives,etc.The distributed blocking flow shop scheduling problem(DBFSP)has more practical application value because it considers the production scenario of distributed flow shop without buffer zone.Based on the actual production scheduling of precast components in prefabricated buildings,this thesis firstly combines the grouping of jobs,assembly stage and energy consumption objective,refines the complex DBFSPs,analyzes the characteristics of the problems,and establishes a mixed integer linear programming(MILP)model.Simultaneously,the swarm intelligence multi-objective evolutionary algorithms are used and improved,based on which innovative empirical strategies are designed to achieve problem solving.Finally,experimental analysis verifies the effectiveness and stability of the proposed algorithms.The thesis mainly includes the following research contents:(1)Starting from the theoretical research,aiming at the energy-efficient DBFSP with assembly stage(EEDABFSP),the MILP model is established to minimize the makespan and total energy consumption,and an improved non-dominated genetic algorithm II(INSGA-II)is designed.The algorithm encodes the feasible solution into a one-dimensional vector containing factory allocation,operation sequence and speed setting.The quality and diversity of the initial population are improved through the initialization scheme based on distributed assembly attributes and the slowest allowable speed criterion.A new crossover operator based on Pareto solution set is adopted,and four mutation operators corresponding to different initialization strategies are designed to enhance the exploration ability of the algorithm.Furthermore,the distributed local search strategies are integrated to expand the mining ability of the algorithm and further achieve the balance between global and local search.(2)In the scheduling of precast components production,a MILP model considering makespan and energy consumption is established for DBFSP with carryover sequence-dependent setup time and considering job grouping(DPGFSP-BCT),and a two-stage coevolutionary algorithm(TS-CCEA)is proposed to solve it.In TS-CCEA,two acceleration rules are designed to save the CPU time,and different initialization methods are used to initialize the initial job population,group population and archive population respectively.In the first stage,mining the specific knowledge of the "classification of solutions",using the variable neighborhood search(VNS)algorithm and the iterative greedy(IG)algorithm to optimize the group population and the job population respectively,and taking the archive population as a bridge to achieve algorithm coevolution.Then,in the second stage,considering the energy consumption,a critical path based speed mutation strategy is proposed to further enhance the exploration ability.Finally,a re-initialization heuristic algorithm is developed to avoid premature convergence.(3)The test instances of precast component production scheduling are constructed to carry out experimental comparison and verification.First,The CPLEX is used to prove the correctness and effectiveness of the model.Then,the orthogonal matrix is used to adjust the algorithm parameters to find the optimal combination.Using the evaluation indicators of hypervolume,inverted generational distance and relative percentage increase,compare the performance of all strategies within the algorithm,and compare the performance other algorithms and the proposed algorithm in the same instance to verify the feasibility and effectiveness of the proposed INSGAII and TS-CCEA,and find the optimal non-dominated solution set within a limited range.