Research On Intelligent Packing Optimization Method In The Manufacturing Of Plate Products

The plate product is made with plate as the main raw materials, and characterized by plane processing as the main manufacturing method, which formed from plate parts with lightweight assembly relations. The plate product is quite common in everyday life including plate furniture, glass deep processing, native building materials, paper products, and printing industry. As the first procedure in the plate product manufacturing, cutting stock optimization is a combinatorial optimization problem, which is of great academic and practical significance. However, manufacturing enterprises of the plate product will be faced with new and rapidly changeable production characteristic driven by highly personalized customers'demands,such as large-scale multiple size, irregular slab, special process requirements, procedure optimization and process coupling optimization.The dissertation focuses on cutting stock and packing optimization problem for the plate product manufacturing process. Taken the manufacturing constraints and process constraints of production process into considerations,it found the heuristic role of geometry approximation relation for the plate product, made use of the mapping relationship betweencharacteristics approximation and process similarity, and explored rapid packing optimization method based on manufacturing procedure and processes dependency. The specific research works are as follows:(1) The techniques system of the grouping method was constructed for solving packing problem. The key techniques of group method includes rapid group construction method by meeting the control depth and hierarchy, group storage organization based on deque, rapid group recommendation algorithm based on fitness. The hybrid construction heuristic packing search model based on group was proposed.(2) A novel two-stage hybrid heuristic algorithm for one-dimension packing problem was proposed. At the first stage, it can quickly generate feasible cutting modes by using greedy and progressive idea; At the second stage, it can translate one-dimension packing problem into a set covering problem, develop the corresponding integer programming model,reduce the problem space by using divide and conquer technique, and integrate CPLEX optimization solver for fast-solving the problems of both modes.(3) To solve the two-dimensional rectangular single bin-size bin packing problem with guillotime constraint, the dissertation proposed a heuristic search algorithm which is called based-group bigitem smallitem divide-and-conquer best-fit heuristic algorithm (BSDBF for short). The algorithm performs a dual recursive search to realize layout-cutting procedure with 2-groups, and integrates bigitem smallitem divide-and-conquer strategy and rapid group recommendation strategy. The initial packing solution is achieved greedily for all bins with binary search, and a postprocessing algorithm is used to obtain a better repacking solution based on heuristic plate splitting strategy. The BSDBF algorithm has characters such as global search ability, constructivity and deterministic, and the optimal solution can be obtained again.(4) To solve the two-dimensional rectangular multiple bin-size bin packing problem, the dissertation proposed a material group-driven heuristic packing approach (MGDHPA for short). Depending on the natural dimension gaps of material group utilization ratio,MGDHPA uses some operations to choose combining bins to zip packing space, including Construct, Recommend, Choose and Replace. By global utilization ratio feedback,it proposed a macro-micro combinational optimization mechanism to coordinate packing process based on plate group and item group. A hybrid repacking approach was proposed to promote optimal solution according the analysis of the results, including three main strategies, namely, the last material repacking, small-scale repacking with replacement of material group, and large-scale repacking with well and inferior set. Combining with MGDHPA, a 3NE level packing algorithm was proposed to apply in three-stage level packing to achieve efficient guillotine cutting result.(5) Aiming at the coupling optimization problem of the cutting, dispatch and loading existed in the plate production, the dissertation proposed a integrated optimization method based on loading dispatch of the buffer warehouse. Based on the proposed 3NE level loading algorithm using k-groups, it can transform the coupling relationship of cutting pattern and loading pattern into pipeline sequence problems to realize the strategic decoupling. By using one-dimension packing optimization methed, the loading problem of the buffer warehouse with sequence constraint was solved based on the idea of layered iteration to realize the computing decoupling.The main innovative points of this dissertation are given as follows: first, it proposed grouping method based on geometrical similarity to improve packing granularity for two-dimension cutting-stock problem, and formed a macro-micro combinational optimizaiton machanism with item groups and combining bins to zip the solve space. Second,it proposed an efficient bigitem smallitem divide-and-conquer best-fit packing algorithm for the guillotine rectangular packing problem, which can be used as the algorithm kernel applying to other solving algorithms. Third, it proposed a new algorithm—three strategies of hybrid construction repacking algorithm which is effective to improve secondary repacking utilization of multi-sized plates. Last, aming at coupling phenomenon of several discrete optimization problems with centered on packing, it developed a mixed decoulping methed based on domain-dependent and combination of calculation and strategy.By the analysis of algorithm performance with many international benchmarks, theresults show the performance of these algorithms is effective. Aiming at combined packing optimization problem of hollow glass manufacturing, the actual test cases of the enterprise prove the engineering practicability of the proposed methods. The relevant methods of the research can provide efficient services of intelligent packing optimization and supports of production decision.