Finishing Line Scheduling in the steel industry

IBM Journal of Research and Development, Sep-Nov 2004 by Okano, H, Davenport, A J, Trumbo, M, Reddy, C, Et al

A new solution for large-scale scheduling in the steelmaking industry, called Finishing Line Scheduling (FLS), is described. FLS in a major steel mill is a task to create production campaigns (specific production runs) for steel coils on four continuous processes for a one-month horizon. Two process flows are involved in FLS, and the balancing of the two process flows requires resolving conflicts of due dates. There are also various constraints along the timeline for each process with respect to sequences of campaigns and coils. The two types of constraints-along process flows and timelines-make the FLS problem very complex. We have developed a high-performance solution for this problem as follows: Input coils are clustered by two clustering algorithms to reduce the complexity and size of the problem. Campaigns are created for each process from downstream to upstream processes, while propagating upward the process timings of the clusters. Timing inconsistencies along the process flows are then repaired by scheduling downward. Finally, coils are sequenced within each campaign. The FLS system enabled a steel mill to expand its scheduling horizon from a few days to one month, and to improve decision frequency from monthly to daily.

1. Introduction

In a major steel mill in Japan, production scheduling for steel sheet products, also called coil products, was conducted by skilled human experts, and the scheduling horizon was limited to a few days. The steel mill had finished automating the scheduling of primary steelmaking, which covers the upstream processes of coil production, but their problem was how to determine accurate due dates for slabs for a longer horizon in order to utilize the automated scheduler more efficiently. Slabs created by the upstream processes are transformed to hot coils at a hot strip mill (HSM) and to cold coils at a cold mill (CM), and are finished with annealing and galvanizing processes (Figure 1). The whole scheduling task for coil production after the HSM, from the CM onward, is called Finishing Line Scheduling (FLS). In order to determine accurate due dates for slabs, therefore, the HSM and the finishing lines must be scheduled for a longer scheduling horizon. The steel mill first examined the feasibility of the existing capacity planning tools for the FLS and found them not applicable. The steel mill then created a special-purpose scheduling system, called the FLS system, with the support of an IBM team including the authors.

A schedule for the finishing lines consists of production campaigns1 on each process and a coil sequence within each campaign. The schedule involves horizontal flows of production campaigns along timelines on each process and vertical flows of coils from upstream to downstream processes (Figure 2). For simplicity, in Figure 2, arrows are drawn for only a few pairs of coils, and only two campaign sequences are drawn for each process. There are actually more arrows, and there may be more than two machines of the same type, called lines, for each process. Each coil has several properties: width, thickness, length, campaign type, release date, due date, priority, grade, and so on. Among the coil properties, those other than release and due dates, priority, and grade are all process-dependent. Note that campaign types on one process are not related to those on other processes. As shown in Figure 2, therefore, two coils in the same campaign on CM may (usually) be assigned to different campaigns on lower stream processes. The number of coils assigned to a campaign is typically about 50 to 500. The task of sequencing the coils assigned to each campaign taking into account the several constraints between coils is called the sequencing (Figure 3). In this task, the vertical relationships should be taken into account, which means that a coil must be processed after it has been through any preceding prerequisite processes. To create campaigns for a particular process, it is usually necessary to create several campaigns of the same type in order to meet the due dates and satisfy the constraints on the minimum and maximum campaign sizes. Therefore, creation of campaigns involves the partitioning of coils into several campaigns with the same type so that the sequencing problems for those campaigns become feasible. Whether or not an instance of the sequencing problem has a feasible solution, in this paper, is called sequenceability. The task of creating campaigns for each process and partitioning coils into campaigns so that every campaign is sequenceable is called campaign allocation. The vertical relationships should be taken into account in this task as well. There are several types of campaigns, and some campaign types require setups for rollers, galvanizing liquid, and so on. Specific campaign transitions minimizing setup costs are preferred, and some campaign transitions also require minimizing setup times. Taking into account the setup costs and times, the problem requires observing the preferable sequences of campaign types, called the campaign templates. The problem also requires creating specific types of campaigns within specific time frames in each month or in each week. Such a time frame is called a chance, and the types of campaigns that must be scheduled in chances are called the chance campaigns. Note that the campaign templates, chances, and minimum and maximum campaign sizes render the allocation problem so difficult that simple capacity-planning tools which do not consider relationships between campaigns along the timelines cannot be used.