World-class computational scientists at Lawrence Livermore National Laboratory[1] (LLNL) are partnering with U.S. Steel to improve its hot-strip-mill operations.

Lawrence Livermore National Laboratory (LLNL) and U.S. Steel are working in partnership through the High-Performance Computing for Manufacturing (HPC4Mfg) program to apply high-performance computing (HPC), modeling and simulation to improve industrial operations and reduce waste at hot-strip mills.

The HPC4Mfg program, operated by the Department of Energy (DOE) Advanced Manufacturing Office, creates targeted partnerships among competitively selected U.S. manufacturers and a national lab or labs to solve intractable industrial problems. Manufacturers receive access to tap the special capabilities of the national-lab partner, especially in modeling, simulation, data analysis and machine learning. The broad benefits of HPC4Mfg include better energy efficiency, reduced waste, and greater competitiveness and leadership in global technology.

Overview of the hot rolling process

Fig. 1.  Overview of the hot rolling process.  A slab is placed in the reheat furnace and brought to a temperature of ~2300°F. It is then extracted from the furnace and squeezed by a series of rollers into a strip ~0.75 inch thick. The strip is then cooled with water sprays, wound into a coil and allow to cool to room temperature over the next few days.


In a project with U.S. Steel, LLNL computational physicists built models of the hot-rolling process to run on LLNL’s HPC platforms. The models track the steel from reheat-furnace dropout through the subsequent steps of rolling, cooling on the runout table, coiling and, finally, post-rolling cooling (Fig. 1). By simulating changes in temperature and steel deformation, researchers can calculate the evolution of grain structure and predict variations in strength properties in silico without the high cost of running experiments. This data is used to inform decisions about the process and improve yields.


Project Background and Objectives

Today’s demand for thinner and wider advanced high-strength steels (AHSS) presents a significant challenge for steelmakers working within the constraints and production costs of existing rolling mills. The new AHSS automotive-grade steels in development require significantly greater control of steel microstructures, starting early in the manufacturing process, to obtain the unique combination of high strength and high ductility desired.

The manufacture of sheet metal involves many steps. First, molten iron is extracted from iron ore in a blast furnace and transferred to a ladle, where its impurities are removed and elements are added to alter the chemistry. The molten steel is then cooled, cast into large slabs and transferred to a hot-rolling mill.

Slabs in the rolling mill are placed in a reheat furnace and heated to temperatures approximately 2300°F. The hot slab is then extracted from the furnace and squeezed by a series of rollers into a strip roughly 0.075 inch thick. The strip is cooled with water sprays on a runout table, wound into a coil and allowed to cool to room temperature over the course of few days. Finally, the hot-rolled strip is pickled to remove the oxide layer (scale) and cold rolled in a tandem cold mill.

All these steps play a role in the final sheet-metal product, but the key factors that define the strength of hot-rolled steel are the composition of the alloy and the thermomechanical changes that occur during hot rolling. These changes drive the evolution of microstructural properties such as the shape and crystalline structure of the grains, which determine the strength and ductility of a steel sheet.

It is essential to control the thermal and mechanical conditions imposed on the steel during hot rolling, but it is difficult to obtain internal measurements while the steel slabs are being processed. As an alternative, numerical models that predict the thermal, mechanical and microstructural changes in a slab throughout the process are an important workaround.

Computational engineers at LLNL are developing 3-D physics-based simulation models that enable steel engineers to peer inside virtual steel slabs throughout the hot-rolling process. Using data from U.S. Steel hot mills, the LLNL team calibrates the models to generate predictions well-tuned to actual operations in the mill.

In previous work, U.S. Steel created 1-D models to simulate variations in thermal and mechanical conditions throughout the thickness of a slab. LLNL and U.S. Steel are now improving this modeling on many fronts, adding 2-D and 3-D models of varying complexity to simulate different aspects of the hot-mill process. Many of these simulations are small enough to run on a high-powered workstation in minutes or hours, while more sophisticated simulations require a supercomputer with hundreds of CPUs. This allows U.S. Steel to run the lightweight models in-house for day-to-day operations and engage LLNL HPC to run high-fidelity models for deeper study.


Temperature data from a StarCCM+

Fig. 2.  Temperature data from a StarCCM+ simulation of the thermal effects on the slab as it travels from left to right through three rollers. Heat from the slab is transferred to the rollers through contact, lost to the air through convective cooling and transferred to the environment through radiative cooling.


High-Performance Computing for Energy-Innovation Program

The Advanced Manufacturing and Fossil Energy offices in the DOE Office of Energy Efficiency and Renewable Energy promote targeted collaborations among national laboratories and the U.S. manufacturing industry through the High-Performance Computing for Energy Innovation (HPC4EI) program. Led by the Lawrence Livermore National Laboratory, HPC4EI matches high-performance computing experts at participating national labs with industry experts to significantly improve manufacturing and/or advance clean-energy technology by solving hitherto intractable problems. By using HPC for the de-sign of industrial products and processes, U.S. manufacturers reap benefits such as accelerated in-novation, lower energy costs, reduced waste and shorter testing cycles and time to market.

For the latest updates on the HPC4EI Program, please visit


Results and Technical Innovation

The U.S. Steel–LLNL collaboration has produced an impressive body of work in the construction of lightweight and heavyweight models. Early work focused on upgrading U.S. Steel’s in-house 1-D mod-el of the mill (up to the coiling operation) to a full 2-D model, allowing researchers to predict the temperature variation in a steel slab from one edge to the other. Edge-to-edge variation is critical information because the edges tend to cool faster and become brittle and unusable. U.S. Steel used the improved model to evaluate over 100 of its steel-sheet products and obtain guidance on how to roll new AHSS grades with better yields.

Recent modeling work has shifted to the use of commercial simulation tools to track the thermal effects in steel, from the dropout of the slab to post-coil cooling in the warehouse. Two StarCCM+ models are used: the first to simulate thermal changes through the rolling and water-cooling steps (Fig. 2) and the second for the long-term cooling of the coil.

Using these models, a series of simulations was completed at LLNL on the effects of uneven heating in the reheating furnace and of masking to change the spray patterns of the water used for cooling after rolling. A full simulation of both models can be run on a workstation in less than one day and will soon be used in-house by U.S. Steel to continue studying temperature variations. These simulations have powerfully contributed to the understanding and control of the hot-mill process.

LLNL is complementing these efforts by applying more-sophisticated models to the hot-rolling process using ALE3D on HPC clusters to run detailed simulations of the structural interactions between slabs and rolls and the nature of sheet deformation as it travels through the roll bite (Fig. 3). These simulations, which require hundreds of processors on LLNL supercomputers to complete, can be used to study problems that may lead to catastrophic failure, such as when a sheet jams and production must be halted.

Data from an HPC simulation

Fig. 3.  Data from an HPC simulation of a steel slab deforming under rolls. As the hot slab is de-formed, thermal conduction is causing heat to flow into the surface of the rolls. This simulation ran on 1024 CPUs of LLNL Cab supercomputer.


Technology Outlook and Summary

Partnering with LLNL computational scientists and harnessing the power of HPC simulation will enable U.S. Steel to improve its rolling operations and promote further innovation in new grades of AHSS sheet. The models developed by the partners span the entire hot-mill process – from dropout through deformation, water cooling, coiling and passive cooling in the warehouse.

U.S. Steel is using some of these models in-house to improve its hot-mill process and avoid edge cracking in new grades of AHSS with fewer expensive experiments being carried out. The team is working to enable more U.S. Steel use of these models, and employing sophisticated HPC models to study rolling-mill jams and other failures is under discussion.

For more information: Contact Aaron Fisher, Head of Numerical Analysis and Simulations, Center for Applied Scientific Computing, at Lawrence Livermore National Laboratory, Livermore, Calif.; e-mail:



  1. Prepared by LLNL under Contract DE-AC52-07NA27344.


All graphics provided by the author.