# New HPC4EI Project Aims to Cut Energy Costs of Painting Cars

September 23, 2020

Contact: [email protected]

Unless you’re in the automotive business, you might not know that painting a car is a surprisingly complex and energy-intensive process. It is by far the largest use of energy in an auto manufacturing plant, averaging 50-70 percent of a plant’s total energy costs. And although automated, it is laborious and time-consuming as well. Cars are painted in a specialized “spray booth” using a multi-step process – including preparation, primer, basecoat, and clearcoat – in which each layer is applied one at a time and then sent through an oven the length of a football field to cure before the next layer can be applied and “oven-baked.” All to create the smooth, shiny, durable coating – with a thickness about that of a human hair – that makes a car’s exterior so attractive.

So for years, the coating industry has been working to develop energy and environmentally friendly improvements to the automotive painting process, such as removing paint layers and cure steps, lowering the temperature of the curing process, and finding new coating substrates. Toward this end, researchers from the Mathematics Group within the Computational Research Division (CRD) at Lawrence Berkeley National Laboratory (Berkeley Lab) are partnering with PPG Industries – one of the world’s largest paint manufacturers – on a new High Performance Computing for Energy Innovation (HPC4EI) project that aims to couple advanced mathematics with HPC resources to model the paint drying process and guide the development of new energy-efficient coatings systems for the automotive industry. The project was announced in June by the U.S. Department of Energy.

This is not the first time the two organizations have worked together. Since joining forces in 2016 through the HPC4Mfg program (now part of HPC4EI), Berkeley Lab and PPG have developed and applied a numerical framework to study and improve rotary bell atomization, another key step in the automotive painting process. Supported by the DOE Office of Science Advanced Scientific Computing Research (ASCR) program and ASCR Leadership Computing Challenge resources, that project made use of multi-million CPU hours to run simulations at the National Energy Research Scientific Computing Center and delivered brand new insights into atomization.

With this new project, the collaborators are focusing on modeling the complex physics that contribute to paint flow and leveling as the coatings dry and cure, with a goal of finding ways to reduce energy consumption in the process.

“Our group in CRD develops new algorithms to tackle rather challenging problems involving fluid-interface dynamics, such as the atomization problem in our earlier work with PPG,” said Robert Saye, a member of CRD’s Mathematics Group who is co-PI on the project, along with James Sethian, head of the Mathematics Group, and Luke Corcos, a mathematics graduate student at UC Berkeley. “We are now working with the same PPG team but looking at very different kinds of physics. It’s not spray painting anymore – actually, ‘watching paint dry’ is far more complex than that sounds.”

## New Mathematical Solutions

Paint must flow and level to form a smooth film that provides an acceptable appearance and protective barrier. But the final appearance is impacted by many variables, and achieving an acceptable appearance is one of the most difficult challenges in developing energy-efficient coatings processes.

With these challenges in mind, two goals for this project are:

- To look for new ways to reduce the temperature it takes to set the paint, and thereby the amount of energy the ovens use
- To study the potential for “co-curing” – that is, painting two layers at once and then setting them at the same time.

“Typically when you come up with a new paint that has different pigments, colors, composition, etc., there is no good way to test it besides firing up an oven and trying things out, varying all sorts of parameters – it’s a lot of trial and error, and this takes time and energy,” Saye explained. “One of the grand goals of this project is to develop a mathematical model, and accompanying algorithmic simulation that could make these predictions quickly and more universally.”

The insights gained from this research are expected to accelerate the introduction of new multi-layer coatings systems that can be co-cured in a single, lower-temperature bake, reducing paint line energy consumption for automotive original equipment manufacturers by up to 30 percent. Beyond the energy savings achieved from fewer curing steps and faster process times, “our research will provide a foundation for future models for water-based coatings and lighter-weight vehicle substrates,” noted Xinyu Lu, a PPG development engineer in automotive OEM coatings, in a PPG press release announcing the project.

This work will utilize some of the numerical methods – such as highly accurate flow solvers and interface tracking techniques – and components of the mathematical framework developed by Saye and his CRD colleagues as part of the initial PPG collaboration. But it will also require the team to invent new algorithms and mathematics.

“Tackling the interlocking physics at play – including interface dynamics, fluid mechanics, heating, curing, impurities, Marangoni stresses, and evaporation – is a formidable challenge,” Sethian said.

Part of the fun, Saye added, is figuring out the right math and algorithms to make the framework handle complex 3D simulations and be viable on supercomputers.

“The goal is to develop a computational model that is capable of modeling coating flow, where you have one or two films of a solute-solvent mixture, with different ‘knobs’ turned on so we can model how the liquids flow, evaporate, and dry out as the temperature of the oven changes,” Saye said. “This project is a combination of part execution of techniques we’ve recently developed and part research of brand new methods. Implementing these new algorithms to take advantage of high performance computing will also be key”.

The initial project timeframe is 18 months – but, if successful, it could be extended an additional 12 months.

“One of the really nice things about these HPC4Mfg projects,” Saye said, “is that when you have a successful collaboration, it ends up using and developing a lot of new technologies – not only computational modeling but brand new numerical methods that can be used in different types of simulations.”

“The mathematical and numerical technology developed here will have considerable applications to other industrial and engineering challenges aimed at optimizing complex manufacturing processes,” Sethian added.