Improving Hot Mill Operations via Optimization

Introduction

In 2015 a collaborative research and development project was undertaken between databright and ArcelorMittal Dofasco in Hamilton, Ontario. Dofasco was interested in developing mathematical optimization models and solution architecture for scheduling and sequencing of their hot rolling mill operations.

Over the course of two years research was conducted, models and software applications were developed, and real production trials were conducted at Dofasco’s hot rolling mill.

The result of this project was a set of novel developments in the field as well as a reoccurring annual benefit to Dofasco conservatively estimated at $1.2M CAD per year.

The follow provides a very high-level description of this work. If you're interested in additional technical details please feel free to contact us as we have a number of documents and presentations which review the mathematical models, solution algorithms, software, as well as performance data.

Background

The steel industry has become a challenging place in the last decade due to increased competition, more demanding customers, and knowledge loss due to an aging workforce. As a result, Dofasco understands the need for continuous improvement and recognizes that advanced solutions such as rigorous mathematical optimization can help facilitate the next generate of improvements within the organization.

Scheduling of production within an integrated steel plant such as Dofasco is still very much a manual process, facilitated by employees, and guided by heuristic rules developed over decades of experience. While the current scheduling team at Dofasco operates at a high level of performance, humans simply cannot be expected to conduct the kind of rigorous optimization required for truly optimal production.

Dofasco engaged databright to conduct a traditional research and development program into optimal scheduling of hot rolling operations at the firms Hamilton plant. This involved the following stages:

  1. Research into the historical and state of the art within this technical domain.
  2. Scoping of the Dofasco-specific problem.
  3. Development of mathematical models for scheduling and sequencing optimization.
  4. Development of solution architecture and software applications.
  5. Conducting production trials on-site using the developed software solutions.

Solution

To begin, research was conducted into the state of the field in the broader space of optimization within integrated steel plants in general. It was found that since the 1970’s there has been a moderate level of both academic and industrial research within the domain. Often the research literature presented more general concepts and high level ideas rather than detailed formulations or production implementations. Overall it was found that there was very limited use of true mathematical optimization within production hot rolling of steel.

With some background in the current state of the art databright set out with Dofasco to develop a solution unique to the company’s hot rolling operations. Over the course of a year databright consulted with subject matter experts within Dofasco to understand and scope the problem. Once the problem was fully understood we set about developing a solution which resulted in a number of novel developments within the field and a final software application for optimal scheduling and sequencing of Dofasco’s hot rolling operations.

The resultant solution architecture allowed Dofasco to schedule and sequence their hot rolling operations for a time horizon which could be defined and adapted by the scheduling team. A series of production trials were conducted, where the hot rolling operation was scheduled and sequenced using the software developed by databright. From these trials slabs of steel were scheduled and sequenced into production lineups, coils of steel were produced within the hot rolling mill, and finally these coils were shipped to customers or further downstream operations.

Various key performance metrics were tracked in terms of scheduling performance, as well as metrics related to the optimization models and software such as computation time. In many cases, an improvement of 50% in scheduling performance was noted. Additionally, the solution architecture facilitated extremely fast computation time for the optimization models and algorithms allowing for ease of use by the Dofasco scheduling team; this helped to build great trust in the solution and associated software.

Results & Recommendations

Many discoveries and conclusions were made through the course of this collaborative research and development project.

Key findings include:

  • Discovery of limitations from simple intra-lineup sequencing.
  • The novel concept and theory of width-group-based horizon scheduling for hot rolling operations within an integrated steel plant was formalized.
  • A solution architecture which allows for efficient and easy to use application software was developed. This is highly valuable when building trust for the solution software within the scheduling team.
  • The proposed solution reliable outperforms the human scheduler. Often scheduling performance metrics showed roughly 50% improvements.
  • Monetary benefits were calculated conservatively at $1.2M CAD per year through the use of the developed solution.

We continue to work with ArcelorMittal Dofasco on this project as the company’s requirements evolve and they expand their application of advanced optimization solutions beyond hot rolling.

Again, if you’re interested in additional technical details please feel free to get in-touch with us. We’d be more than happy to share these details with you!