Research On Slab Allocation Problems Arising In Parallel Hot Rolling Production Lines

Slabs produced from continuous casting stage are reheated and rolled into hot coils in the hot rolling production line. Open-order slabs refer to the extra slabs that are produced to satisfy the batch production mode of large steel making facilities but not required by any customer orders at present. Open-order slab allocation is to allocate the open-order slabs to unfulfilled customer orders by considering the characteristics of slabs and the requirements of customer orders, such as steel grade, width, length, weight and so on. Timely and reasonably allocating of open-order slabs will directly affect the resource utilization and the order's fulfillment rate, hence slab allocation problems have received increasing attention by academics and the steel enterprises.According to above description, from the perspective of production planning, slab allocation problems with different characteristics including dynamic arrival of slabs, diversification of allocation patterns, stochastic yields of open-order slabs, and stochastic demands of customer order are studied. For each of above problems, modelling and optimization methods are tailored designed. In addition, in order to evaluate the performance of slab allocation problem in operations management, a model predictive control based approach is proposed from the perspective of production management. Furthermore, from the perspective of practical application, two decision support systems for slab allocation in parallel hot rolling production lines and operations management performance are developed. The detailed contents of this paper are as follows.1) We investigate the dynamic slab allocation problem for parallel hot rolling production lines, considering dynamic arrival of slabs and three kinds of allocation patterns such as allocation to the customer order, allocation to the potential order and allocation to the self-designed order. Using the allocation pattern and allocation relationship with order of each slab as decision variable, considering the practical constrains such as order intergrity and delivery constraint, safety inventory constraint, we formulate the problem as a binary integer programming model. The objective function is to minimize the trade-off between matching cost, inventory holding cost, over-fulfillment discount, under-fulfillment penalty, and setup cost caused by the self-designed order allocation. To solve the problem efficiently and effectively, we design a three-gene structure coding strategy based on which an improved scatter search algorithm is developed. In our algorithm, a local search mechanism based on multiple neighborhoods and a random perturbations strategy are introduced to improve the depth and breadth of searching capabilities. Computational results on the randomly generated instances show that the proposed scatter search algorithm can get near-optimal solution in a reasonable CPU time compared with the commercial optimization software such as IBM ILOG CPLEX. Computational results on the practical data show that the proposed algorithm outperforms the manual planning method which the steel company usually used.2) We investigate the slab allocation time determination problem for the self-designed orders, which is to determine the maximum waiting time (after this time slabs have to be immediately allocated to self-designed orders) for each difficult allocating slab before the potential order appears. For this problem, we use the history information of slabs and orders as input, and formulate a best allocating time parameter prediction model by using a least squares support vector machine based data analytics method. To improve the prediction accuracy, a scatter search algorithm is proposed to optimize the parameters of data analytics model. Computational results show that the least squares support vector machine based prediction model in which the parameters are improved by scatter search algorithm has the root mean square error of 2.40%.3) We investigate stochastic slab allocation problem considering the uncertain information of open-order slabs yields and customer demands under a multiple stages case. It can overcome the limitation of the single-stage and deterministic slab allocation problem that only considers short-term benefits but give up long-term goals. Using the amount of slabs assigned to orders as decision variables, considering the inventory balance constraint, production and inventory capacity constraint and so on, we model uncertainties using a scenario tree based on which a stochastic mixed integer programming model is formulated. The objective is to minimize the total expected value of production and inventory cost. Since the large scale scenario tree based model is difficult to be solved, a model dimension reduction approach based on scenarios aggregation is proposed. A scatter search algorithm with directed local search based on follow-up technique is proposed to solve the reformulated model. Computational results show that the proposed algorithm can get the optimal solution for most of small and medium scale instances, and is superior to the commercial optimization software such as IBM ILOG CPLEX for the large scale instances both on solution quality and computation time.4) We develop the decision support system for slab allocation in parallel hot rolling production lines. Based on the detailed requirements analysis for the slab allocation planning in parallel hot rolling lines as well as the proposed model and algorithm, we develop a decision support system for slab allocation. The system includes data management module, slab allocation planning module, parameter setting module and so on. Moreover, a friendly human-machine interactive interface is also included in the system such that the planners' preferences and personalized management requirements can be expressed flexibly. The system has been implemented and is actively used in a large domestic steel company. By using the system, the open-order slab utilization and the order's fulfillment rate are improved, the inventory and logistics cost are reduced, and the planning work efficiency is improved.5) We develop the operations management performance system to evaluate the performance of slab allocation in operations management. Through the analysis of the relationship between allocation problem and operations management performance, determine the key performance indicators for slab allocation, production materials, and orders in iron and steel production process. In this paper, using the insights of model predictive control, we make analysis for this problem based on the key performance indicators of slab utilization. Based on the idea of performance management, we develop the operations management performance system to keep trace of the materials and orders on the plant-wide process. The system includes data management module, performance management module and so on. The system has been implemented in a large domestic steel company. It can accurately and timely reflect the work state of production process, such that the planners can adjust the working plan at any time to improve the corresponding performance indicator. The system can also provide a decision support function to the managers from the global view by statistical analysis of history data.