A Matheuristic for Scheduling the Steelmaking-Continuous Casting Process with Multiple Refining Stages

Abstract

This thesis presents a matheuristic for scheduling the steelmaking – continuous casting (SCC) process with multiple refining stages. The SCC process is a bottleneck in the steel production, so an effective scheduling for the SCC process reduces costs and allow the just-in-time delivery. To increase productivity of this bottleneck process, some researches have been conducted over the last few decades. The proposed methodologies, however, can hardly reflect the complexity of real-world industry. Therefore we have considered more general cases than the previous studies and suggest a simply extendable and effective algorithm for the SCC process. The SCC process consists of mainly three stages; steelmaking, refining and continuous casting. For the refining stage, it has multiple ones. This problem can be modeled as a flexible flowshop problem with some special requirements which make it more difficult. A job unit is called a ‘charge’ and the charges of the SCC process go through the steelmaking, refining, and continuous casting stage, but they can skip some refining stages by different process plans. It also considers minimal and maximal time lag, which correspond to the transportation time and the waiting time limit, respectively. At the continuous casting stage, some charges in a serial batch should be continuously processed. This batch unit is called a ‘cast’ and each cast has a setup time. To solve this problem in reasonable time, we propose an algorithm consisting of three phases; iterative MIP, destruction and construction, and MIP improvement. This methodology is all based on a mixed integer programming (MIP) model. An initial solution is obtained by iterative MIP which sequences the casts and constructs the schedule of them one by one. We get a better solution by destructing some partial schedules from the initial one and reconstructing them. Moreover, the solution is improved by MIP improvement phase. Our objective is to minimize the total waiting time, earliness and tardiness, taking into account the cost of reheating process and line balancing. Lastly, numerical results with optimality gap are illustrated to validate our algorithm.