Time-space Modeling And Convex Optimization Methods For Steel Production And Logistics Scheduling

Steel production scheduling is to determine the combination,assignment and production time of the objects on production equipment of each stage in the transformation process of physical,chemical and mechanical properties from raw materials to finished products;while logistics scheduling is to determine the allocation of the object the logistics equipment,timing and locations during the operations of accessing,handling and transportation.Scientific production and logistics scheduling can efficiently improve the utilization of large-scale production and logistics equipment,reduce work-in-process inventory,and enhance the intelligent level of iron and steel enterprises.As production and logistics scheduling decisions contain a large number of integer assignment and sequencing variables,and the need to meet multiple management objectives and complex production process requirements simultaneously,making accurate formulation of steel production and logistics scheduling becomes internationally challenging scientific problems.In engineering,the quality of modeling directly affects the performance of optimization,which makes the modeling of production and logistics scheduling has important practical significance.In this dissertation,the problems of production and logistics scheduling in the practical operations management of steel manufacturing system are studied,and the time-space network modeling methods and the convex optimization methods are proposed.For the production scheduling problems such as hybrid flow-shop scheduling,slab-order allocation problem considering logistics cost,and slab-order robust reallocation problem,time-space network modeling,convex relaxation,and hybrid Lagrangian relaxation and convex optimization algorithm are proposed,respectively;For the logistics scheduling problems such as crane scheduling in a coil warehouse and products re-deposit scheduling,time-space network modeling,approximate dynamic programming,and hybrid Lagrangian relaxation and convex optimization algorithm are proposed,respectively.The main research contents are summarized as follows:(1)An efficient time-space network modeling method is proposed for the hybrid flow-shop scheduling problem of steel manufacturing system.The method constructs a network by discretizing time and space into a grid,such that the generalized nodes in the network represent the allocation of the jobs on the machine,and the arcs in the network represent the connection of the jobs between two adjacent production stages.Two kinds of dimension reduction strategies based on starting/finishing time and waiting time of the jobs are proposed to accelerate solving the model.The computational experiments show that the time-space network modeling method outperforms existing modeling methods on solving efficiency and solution quality.(2)The slab-order allocation problem considering logistics cost is derived from the manufacturing management process of iron and steel enterprise,which needs to allocate the open-order slabs produced from steelmaking stage to customer orders under the matching conditions,so that the slab utilization,slab retrieval cost and customers satisfaction are optimized.Since the 0-1 integer quadratic programming model formulated for this problem is difficult to be solved by regular optimization methods,a semi-definite relaxation method is designed to obtain a lower bound of the problem,and a convex optimization based heuristic algorithm is developed to obtain near-optimal solution of the problem.The effectiveness of the proposed method is verified by computational experiments on practical production data.(3)The slab-order robust reallocation problem is derived from the hot-rolling production management of iron and steel enterprises.The problem is to re-optimize the original allocation between slabs and orders due to the deviations of actual produced slabs in the quality,weight and size from the design.A robust optimization modeling method is proposed to deal with the uncertainties of matching parameters,the robustness of the parameters is described as a ellipsoid set.Since the model can not be directly solved,it is equivalently transformed into a mixed integer second-order cone programming through mathematical transformation,so that it can be directly solved by the mainstream optimization software CPLEX for small-scale problems,two classes of valid inequalities are proposed to accelerate solving the model.A hybrid Lagrangian relaxation and second order cone programming algorithm is proposed to solve the large-scale problems.Experimental results show that the proposed algorithm performs better than the mainstream software CPLEX.Based on the proposed model and algorithm,the slab-order reallocation decision support system is developed,which has increased the orders' completion rate,and reduced slab trim-loss and order surplus.(4)The crane scheduling problem is derived from the operations management of steel coil warehouses.The problem needs to simultaneously determine the sequence of handling coil storage/retrieval and shuffling requests as well as the positions to which the coils are moved so that the crane's efficiency is optimized.For the complex process requirements,logistics logic and time-space coupling characteristics of the problem,an event-based continuous time-space network model is proposed,where each node represents a location in the warehouse at the end of a scheduling stage,and each arc indicates a crane's move between two locations in a stage.An approximate dynamic programming algorithm based on optimal assignments in a bipartite network with cuts is designed by exploiting the problem structure to solve large-sized instances.Computational results show that the proposed model is more efficient than traditional modeling methods,and the approximate dynamic programming algorithm is more efficient than the mainstream optimization software CPLEX.(5)The products re-deposit planning problem is derived from the logistical process of iron and steel enterprises.The products that produced from the final process and stored in the terminal warehouse need to be transferred to the finished products warehouses for logistics distribution.The problem needs to determine the time and location for the products re-deposit to meet the capacity of transportation tools and logistics equilibrium.The problem is formulated as a nonlinear integer programming model,the nonlinear objective for logistics equilibrium is equivalently transformed into a mixed integer second-order cone constrant.A hybrid Lagrangian decomposition and second order cone programming algorithm is proposed,and two classes of knapsack valid inequalities are proposed to accelerate solving the subproblems.Experimental results show that the proposed algorithm outperforms the mainstream optimization software in terms of solution quality and solving efficiency.