Accelerating clean energy manufacturing with high-performance computing

Lawrence Livermore has once again drawn upon its high-performance computing (HPC) expertise to enhance industrial practices in the U.S., with an eye toward advancing the nation’s economic competitiveness. This time, Livermore is leading a program, called HPC for Manufacturing (HPC4Mfg), that pairs researchers from several national laboratories and academia with manufacturing companies to deliver solutions that maximize their production environment. HPC4Mfg is funded by the Department of Energy (DOE) Energy Efficiency and Renewable Energy’s Advanced Manufacturing Office.

HPC4Mfg uses the modeling, simulation, and data analysis know-how of laboratory scientists to help solve various challenges in the manufacturing sector. Project goals vary, but examples include optimizing production processes, assisting in decision making, improving product quality, predicting performance and failure, and speeding up or eliminating testing.

“The problems industry tackles are unique from the problems we typically address in-house,” says Livermore computational scientist Aaron Fisher, the principal investigator for two HPC4Mfg projects. “We support industry by producing custom solutions to their manufacturing challenges. Addressing these challenges introduces Livermore researchers to problem spaces they haven’t yet considered, which creates many opportunities for the cross-fertilization of ideas.”

Fisher’s projects, blast furnace modeling and E-Nugget process modeling, aim to improve the nation’s steel production processes by harnessing the power of HPC. In both projects, Fisher and his collaborators are working to reduce costs to the industry, decrease energy consumption, and improve greenhouse gas emissions. Adds Fisher, “There are large levers in the steel industry influencing climate and energy security.”

The virtual blast furnace project is an integrated HPC modeling, simulation, and visualization effort for steel manufacturing that partners Livermore and Purdue University Calumet’s Center for Innovation through Visualization and Simulation (CIVS). CIVS leverages industry connections forged in the university’s Steel Manufacturing Simulation and Visualization Consortium, to identify ways to improve steel production. Purdue also brought to the table three simulation codes they had developed to model blast furnaces, which Livermore researchers are enhancing to reduce their runtime. Purdue uses these codes to explore the gas flow, combustion, and chemical reactions in these billion-dollar blast-furnace systems, which serve as the backbone for steel production.

Use of these codes has already saved the steel industry an estimated $8.5 million a year for blast furnace processes, as well as reduced downtime by 50 percent, through improved natural gas injection. However, each simulation run took more than a month to complete using Purdue’s computer system, and the blast furnace had to be simulated in three separate parts, increasing the cost and runtime. To mitigate this runtime issue, Livermore’s efforts focused on streamlining the codes using a method called multithreading (a type of parallel processing). Livermore improved the runtime of one of the Purdue blast-furnace codes threefold, effectively reducing the typical runtime for a three-dimensional simulation from a month to only 10 days. Additionally, Livermore researchers utilized their HPC clusters to run hundreds of simulations at the same time. This allowed them to analyze the effects of a wide variety of furnace operating conditions within a reasonable amount of time.

Though the six-month effort ended in January 2016, the team has received follow-on funding to further enhance Purdue’s codes using parallel message passing (another parallel programming technique). Livermore plans to further improve the runtime of the codes so that they can complete a single run in one day on a computer cluster and utilize the faster codes to improve the efficiency of blast furnace operation.

A second project joins Carbontec Energy Corporation with Livermore and Purdue to model and optimize their novel approach for converting iron ore to pig iron nuggets using clean, inexpensive, renewable biomass in place of coal coke. The technique, called the E-Nugget process, is expected to lower U.S. iron production costs, enabling U.S. steel companies to better compete with imported iron. Leveraging Livermore’s state-of-the-art supercomputers and modeling expertise, the multi-institutional team will work to optimize Carbontec’s existing design of the furnace for the E-Nugget process. The project goal is to reduce plant construction costs and increase energy efficiency and productivity of the operation. Livermore plans to run simulations to help Carbontec identify the necessary parameters for scaling up the project, in hopes of reaching a production rate of 100,000 metric tons per year.

While still in its infancy, HPC4Mfg has made great strides towards large-scale innovation. “HPC4Mfg may be less than a year old, but it is executing five pilot projects in various stages of completion, has awarded nine proposals from its inaugural solicitation, and has an active solicitation in progress,” notes Peg Folta, HPC4Mfg director. “Other areas of industry have shown interest in the program, which could be leveraged to broaden laboratory partners, as well as increase the impact to our nation’s economic competitiveness.” Adds Fisher, “In addition to improving the efficiency of large-scale manufacturing, this work strengthens the U.S. as a whole by making the industries we work with more competitive,” says Fisher. “A strong U.S. manufacturing sector is a crucial buffer against disruptions in the world economy.” Livermore will continue to establish new ties with industry partners, pushing the bounds of innovation using its HPC expertise.