Modeling And Differential Evolution Algorithms For The Coil-order Matching Problems

During the steel production process, not only the requirements of varieties, small batches and individuation from customers should be met, but full utilization of the unit equipments should also be ensured. However, the ineffective management results in a certain amount of open-order coils produced without customer orders. Moreover, as a result of the unreasonable production plan and some other factors, coils that have been produced for a certain order may not be the best choice for the original order. Coil-order matching problem is to allocate open-order coils to suitable unfulfilled customer orders which can reduce surplus inventory, and reallocate existing customer-order coils among orders to improve the matching relationships which can increase the order completeness and customer satisfaction levels. Taking the practical production process of a domestic steel enterprise as background, this thesis studies the coil-order matching problem and describes it as a two-stage optimization problem including coil allocation problem and reallocation problem. Mathematical models and differential evolution algorithms are proposed for the problems respectively considering the characteristics of the problems. The main contents are shown as follows:(1) Modeling and solution for the open-order coil allocation problem. Aiming at improving the coil-order matching quality and coil utilization, and reducing the inventory cost, the problem is formulated as an integer programming model considering the specifications of coils and the requirements of orders. Since this problem is NP-hard, which is difficult for optimization software to solve in large scale, considering the characteristics of the problem, an improved differential evolution algorithm is proposed with an integer encoding manner and customized heuristics are designed to obtain good initial solutions. The algorithm adopts an improved mutation operation with remainders and uses greedy repair strategies to repair the infeasible solutions. Computational experiments of different scales verify the effectiveness of the model and algorithm.(2) Modeling and solution for the coil reallocation problem. In the practical production process, there often occurs unsuitable matching relationship between coils and orders. Coil reallocation decisions are made to improve unsatisfactory allocations. In order to improve the coil-order matching quality and customer satisfactory, considering the constraints of maximum capacity of each unit, this problem is formulated as an integer programming model. Small-scale problems are solved by CPLEX, which shows the correctness of the model. For large-scale problem, an improved differential evolution algorithm with the disturbance of random mutation and adaptive parameters is proposed whose initial population is obtained by heuristics algorithm and random generation. The experiment for comparing and performance of the proposed differential evolution algorithm and CPLEX shows that the proposed algorithm can obtain satisfied near-optimal solutions within a short computation time.(3) A computerized decision support system (DSS) was developed embedding the proposed models and algorithms above. It consists of the following functional modules: Login module, Data download module, Data query module, Parameter configuration module, Algorithm running module and Data uploading module. The DSS can generate high quality matching scheme in a short time, which can reduce the production cost, improve the work efficiency and customer satisfaction.