Modeling, Simulation and Visualization Capability for Steel Manufacturing

In this case study, Lawrence Livermore National Laboratory (LLNL) is working in partnership with Purdue University Northwest and steel-industry stakeholders to use their high-performance computing (HPC) modeling, simulation and visualization capabilities to optimize blast furnaces in order to reduce emissions and energy use.

The DOE Advanced Manufacturing Office launched the High Performance Computing for Manufacturing (HPC4Mfg) Program to create targeted partnerships between U.S. manufacturers and the national laboratories. Under this program, competitively selected projects take advantage of the unique scientific expertise at the national laboratories in combination with high-performance computing capability. The industry/lab teams apply modeling, simulation and data analysis to industrial processes and products to solve problems that might otherwise be intractable. Benefits include improving energy efficiency, reducing waste, boosting competitiveness and building global technology leadership.

One of the initial pilot projects was a collaboration between LLNL, the Steel Manufacturing Simulation and Visualization Consortium, and Purdue University Northwest’s Center for Innovation through Visualization and Simulation (CIVS) to explore moving CIVS’s existing blast-furnace model to an HPC platform.

At CIVS, the model is so complex that it can only be run in three separate pieces – representing the three major sections of the blast furnace – and it takes an entire month to complete one run. With an industry cost-share commitment of 22%, we decided to take on the blast-furnace modeling challenge and move this project forward. Now in its second and final year, it is already yielding new insights about blast-furnace operations and has become a showpiece of the program.

Project Background and Objectives

Production of iron and steel is a capital- and energy-intensive manufacturing process. It is the fourth-largest energy-consuming industry in the U.S. Twenty percent or more of the cost of producing a ton of steel is attributed to energy, including substantial amounts of coal, electricity and natural gas required for the manufacturing process.^[1]

The heart of primary steelmaking is the blast furnace, which removes oxygen from iron ore and produces molten iron. The blast furnace is a very large, vertical structure where iron ore, coke and limestone are spread into the top of the furnace while preheated air – and sometimes natural gas and pulverized coal – are injected into the bottom (Fig. 1).

The raw materials require six to eight hours to descend to the bottom of the furnace, during which time the iron oxides are chemically reduced and physically converted into high-carbon liquid iron. The iron is drained from the bottom of the furnace at regular intervals^[2] and transferred to a basic oxygen furnace (BOF), where it is converted to steel. Once a blast furnace is started, it will continuously run for four to 10 years with only short stops to perform planned maintenance.

Because it is very difficult to obtain measurements inside these intensely hot furnaces, numerical models provide an important alternative. Modelers at CIVS developed physics-based advanced simulation and visualization technologies to enable researchers to peer (through virtual reality) inside the dynamic and changing environment of the blast furnace. Using data from scientific computations and real operating conditions, CIVS researchers work with steel-industry partners to solve problems and improve blast-furnace processes.^[3] Without the hundreds of thousands of processors that HPC can offer, however, smaller computer clusters or workstations encounter limits to what they can model, including:

Dynamic processes such as start-up, shutdown or periods of instability are excluded because the model is limited to steady-state.
Simulation of the furnace must be performed in three sections, one at a time.
Phenomenological models are reduced in scale, including those for chemical reactions or skull formation in ladles.^[1]

The HPC4Mfg Program is enabling researchers at Purdue University Northwest to partner with HPC experts at LLNL to improve the virtual blast-furnace models through a two-phase project. Phase I focused on examining and scoping out code improvements, including integrating the different sections of the blast-furnace model and parallelizing them to improve runtime and increase resolution. A second objective was to demonstrate the power of using HPC resources by running a series of parameter studies examining coke consumption under various blast-furnace operating conditions and running simulations of alloy mixing in steel ladles.

Phase II is focused on implementing the improvements to the blast-furnace codes identified in Phase I, such that it is possible to run a 3-D simulation on the entire system in a single day. Additionally, researchers are working with industry stakeholders to determine a set of large-scale parameter studies aimed at improving operational flexibility to reduce cost by lowering coke consumption and overall energy use.

Results and Technical Innovation

Research partners are making significant progress on the Purdue blast-furnace code improvements. Integrating and parallelizing the raceway and shaft blast-furnace codes are reducing the runtime by more than a factor of 1,000. Significant algorithmic changes are being made, and LLNL’s open-source, highly scalable SUNDIALS nonlinear solver is being leveraged to significantly improve the ability of the code to run on large numbers of processors. Recent CFD results based on the improved code are shown in Figure 2a. Figure 2b shows the same results combined with a blast-furnace training simulator previously developed by the CIVS.

Parametric studies with hundreds of input combinations were run in parallel utilizing the shaft code to demonstrate the kinds of analysis that HPC can unlock for blast-furnace operators. Figure 3a shows the coke rate, a variable the industry would like to decrease, as a function of percent-oxygen enrichment. The furnace charge, or burden, of iron ore is fixed, but the graph also shows the effect of different wind rates, or speed at which the gas is pushed through the blast furnace.

Figure 3b shows how the wind rate affects the pressure drop across the shaft, with the area between the dashed horizontal lines representing the pressure range of stable furnace operation. The results are encouraging in that they indicate there is a large region of parameter space that can meet the pressure specifications of the furnace operation for various coke and wind rates.

Simulations of steelmaking processes downstream of the blast furnace were also run. The steel industry is interested in reducing the time it takes to sufficiently mix iron with the alloys used to make different types of steel, but past desktop simulations run at CIVS took weeks to complete. The previous simulation was converted to Star CCM and run on Syrah (a large cluster at LLNL) to determine if the simulation runtime could be reduced significantly while also increasing the resolution to capture finer details.

The results of the scaling study indicate the new simulation runs >1,000 times faster at finer resolutions than the previous CIVS desktop runs and >15x faster than the 40 CPU Peregrine system at Purdue University Northwest (Fig. 4). While the focus of the project’s second phase is improving the blast-furnace code and parameter studies to optimize blast-furnace operations, this scaling study demonstrates manufacturing improvements are also possible for downstream steelmaking processes.

Technology Outlook and Summary

Harnessing the capabilities of HPC will enable CIVS to help the U.S. steel industry become more globally competitive by identifying areas for increased energy and cost savings. The long-term vision for the CIVS blast-furnace model capabilities is to develop an interactive virtual blast furnace (VBF) that combines a comprehensive, integrated, high-fidelity, dynamic, multi-physics model with fast data visualization. The VBF will also be a powerful tool for the steel industry to use in workforce training. In addition, by combining the VBF with additional capabilities (e.g., networked sensors, data interoperability and intelligent automation), a smart manufacturing system could be developed to provide optimized, stable and safe operations, along with superior energy efficiency, product quality and productivity.

The project detailed here is a demonstration of what is possible when manufacturers partner with the HPC experts and resources at the DOE national labs. As of July 2017, the HPC4Mfg Program has funded more than 40 projects, and more are on the way from our Spring 2017 solicitation. A number of these projects are leveraging HPC capabilities to study other high-temperature manufacturing processes, including:

LLNL and Purdue University Northwest are partnering with Carbontec to model and optimize the furnace design for Carbontec’s E-Iron process of converting iron ore to pig-iron nuggets using inexpensive, renewable biomass. A VR model of a production-scale furnace design will be completed in 2017 using data from simulations running a set of parameter studies.
Several projects are under way in microstructural modeling and control for alloy development, some of which are being designed for use in additive manufacturing.
Numerical simulations of high-strength molten-glass fibers drawn through a bushing system are being used to study how local variations can cause fiber breakage. A separate project is developing a glass-furnace model enabling informed, real-time process adjustments.

High Performance Computing for Manufacturing Program

The Advanced Manufacturing Office in the Department of Energy’s Office of Energy Efficiency and Renewable Energy enables targeted collaborations between the national laboratories and the U.S. manufacturing industry through the High Performance Computing for Manufacturing (HPC4Mfg) Program. The program, led by the Lawrence Livermore National Laboratory, partners an HPC expert at one of the participating national labs with an industry expert to address challenges that could significantly improve manufacturing processes and/or result in advancing clean-energy technology. By using HPC in the design of products and industrial processes, U.S. manufacturers can reap such benefits as accelerating innovation, lowering energy costs, shortening testing cycles, reducing waste and cutting the time to market.

For the latest updates on the HPC4Mfg Program, please visit http://hpc4mfg.org.

For more information: Contact David Forrest, technology manager, Advanced Manufacturing Office, at the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy, Washington, DC; tel: 202-586-5725; e-mail: david.forrest@ee.doe.gov; web: https://energy.gov/eere/amo/advanced-manufacturing-office

References