Novelis Leverages HPC Platform and SigOpt to Optimize Product Resiliency and Security
As the world’s largest aluminum sheet flat rolling and recycling company, Novelis has more than 14,500 employees in nine countries. Its business is focused on the production of environmentally responsible non-ferrous products for its customers spanning many industries, including automotive, appliance, beverage and aerospace. For Novelis customers to remain competitive in today’s fast-moving marketplace, reducing the time required for product research and development is critical.
In the past, Novelis relied on in-house CAD design workstations, and trial-and-error approaches, to run simulations and identify the best product designs for prototyping and testing. However, as the designs grew in complexity, the time required to complete them also increased. This reality made legacy workstation-based approaches slow and impractical. For example, a Novelis customer needed a stronger and lighter beverage can. Designing a can may seem like a simple task. In reality, beverage cans are one of the most highly engineered products manufactured today. Small changes to the geometry of the product can result in profound performance improvements. Therefore, identifying the best possible design requires consideration of 26 different parameters through 67 million finite element (FE) calculations. If you use typical FE Analyzer methods for the task, the design cycle could take 150 years!
Credit: Image courtesy of Novelis.
To speed up the iterative process, Novelis turned to simulations using high-performance computing (HPC) cloud instances powered by Intel Xeon Platinum processors and SigOpt’s intelligent experimentation platform. As explained in a recently published article case studyInstead of prototyping and testing ten or more designs in a lab, the new approach can narrow that list down to two or three. As Dr. Vishwanath Hegadekatte, Global Director of R&D for AD and Advanced Modeling at Novelis, said: “With the help of HPC, we can run simulations much faster and reduce the time required for physical prototyping. Also, we can deliver better products to our customers much faster.”
Better, stronger, faster
Typical aluminum cans today weigh 14.5 grams with a wall thickness of 0.21mm. That’s a huge improvement over past years. However, reducing that weight by another gram would save approximately 23,000 metric tons of aluminum per year. Using modern simulation tools, engineers found that shallower curves at the top and bottom of the can improve its force-to-weight ratio to conserve resources. The result of applying HPC technology to the process makes the soft drink can one of the most highly engineered products you will ever have in your hands.
Improve car safety
Another Novelis customer, an automotive company, sought help to improve the safety of their vehicle with a new geometry of aluminum sheet that gradually folds during a crash, without cracking, to reduce the impact on vehicle occupants. Designing an “axial bearing component” like this with traditional FE analysis could take years or even decades. With the SigOpt solution, Novelis completed the task quickly without the previously time-consuming trial and error method. The process also provided a superior result while reducing development costs.
Today’s projects accelerate future design iterations
AI-based modeling and simulation also brought other advantages to Novelis by generating “reusable” data. When Novelis engineers run new math or physics simulations in the future, they will be able to run fewer tests faster while extracting valuable insights. “The main thing that surprised us was that the performance increase we saw with AI was nearly double that of our manual approaches,” Hegadekatte said. While the AI-based modeling simulation has a proven track record of accurate results, the Novelis team still performs a small number of physical experiments to verify the expected performance of the product.
As the examples of beverage cans and automobiles illustrate, modern design processes can deliver better, more cost-effective, and more environmentally friendly results. As HPC and AI technologies advance, it’s exciting to imagine how many familiar objects around us could be improved in unexpected ways. As Hegadekatte noted, “With the help of Intel software and SigOpt, we can explore new design approaches that are impossible for humans.”
About the Author
Rob Johnson spent much of his professional career as a consultant for a Fortune 25 technology company. Rob currently owns Fine Tuning, LLC, a strategic marketing and communications consulting firm based in Portland, Oregon. A lifelong technology, audio, and gadget enthusiast, Rob also writes for TONEAudio magazine, reviewing high-end home audio equipment.