Toyota’s Inverse Design Method for Fuel Cells

Multiphysics simulation works to find material properties, then create a design.

The Toyota Research Institute of North America (TRINA) is using an interesting simulation approach to find new flow field designs for hydrogen fuel cells. Using COMSOL and MATLAB, it is working backwards to optimize the design of fuel cells. Its inverse design method takes the target material properties and creates a design to meet that target. This is leading researchers to look at other applications for this method.

Flow field prototype built from inverse design generation: COMSOL.

Flow field prototype built from inverse design generation: COMSOL.

Toyota has been all-in on the idea of integrating hydrogen into its products for decades. The Toyota Mirai is a commercially available automobile in the small-but-growing segment of electric vehicles. The Mirai, however, is a fuel-cell electric vehicle (FCEV) so it exists in a sub-section of a sub-section of automobiles. Most vehicles don’t need to answer questions about fuel safety in the FAQ section of their marketing data, but the Mirai does because of the public perception of hydrogen tanks in a vehicle. This makes the engineering, design and simulation of these products more important.

The sustainability problems facing the world right now are too big for any one solution to fix. Engineers and researchers are working on multiple designs for multiple problems, all hoping to incrementally make the world a better place—and hydrogen is a part of this future. Open-minded application of design principles will challenge how we think a conventional product should look and the benefits should justify these radical approaches. For instance, this implementation of COMSOL shows that using simulation and multiphysics to reverse design geometry will help optimize fuel cell design and can be transferred to other areas of manufacturing and development.

Toyota and the Vision of a Hydrogen Society

Toyota started the development of a hydrogen fuel cell vehicle way back in 1992, and has been working toward integrating hydrogen into society throughout this process. In 2018 it announced the intention to become a mobility company across several avenues: “Mobility will be seamlessly integrated into essential aspects of society such as healthcare, food and agriculture, energy, finance and education, and be used in various ways that inspire daily happiness and amplify human potential.” This wildly ambitious shift in what we traditionally think of as mobility led Toyota to start development and construction of 175 acres of test space called Woven City. 

One of the big lifestyle shifts that comes from Woven City is the portable hydrogen cartridge. These futuristic gas cans could be used to provide energy to homes, power to drones, charge electronic devices and deliver energy to a new fleet of fuel-cell electric vehicles. Having a clean energy source easily available to power vehicles could change how we think about mobility in cities and beyond.

The Inverse Approach to Design for Fuel Cells and Beyond

To help make Woven City possible, TRINA researchers are using simulation to drive a generative design process using COMSOL and LiveLink for MATLAB.

Electric current is generated when hydrogen and oxygen are flowing through the fuel cell. Flow field plates dictate how the hydrogen and oxygen will flow, and creating the best possible paths for flow can increase the output of the fuel cells. Current designs of the flow plates come from conventional thinking and straightforward design. Parallel flow patterns and serpentine patterns already exist, and some other designs have a spiral or pinned grid pattern.

Creating a new way for the hydrogen and oxygen to flow through the plates required a little unconventional thinking. Fuel cells are generally designed forward: first deciding how the fluid will flow and then making small modifications to the size or reaction surface for optimization.

The TRINA team decided to use an inverse design method, working first on material properties and distribution and leaving the flow field design until later. The inverse design method meant that instead of creating a configuration and then analyzing that design, a required set of inputs and outputs was used to find optimal flow characteristics and the design was built from those characteristics. Multiphysics simulations and iterative designs are used to find the flow path; this method has previously worked on fuel cell design. The drawback was incredibly high computational costs, so the inverse method was not widely used.

However, when the team used a coarser mesh to lower the computational power required for the simulations the method worked more efficiently. Cell modeling and 2D analysis are used together to approximate the 3D space of the fuel cells, and then a time-dependent reaction-diffusion study points toward an optimized flow model. COMSOL LiveLink for MATLAB pulls multiphysics data from COMSOL into an optimization loop in MATLAB with a maximum of 200 iterations.

Working at the molecular level lets the engineers figure out the microstructure required from the system, then move into the required material properties for the working fluid, and finally dictates the structure required by the flow plates. These results look more organic than what most people think of when envisioning flow patterns. After building flow channels around these results validation models are run in the software, and a few physical models were created from the best-developed models.

Where Do We Go From Here?

The inverse design approach used here really punctuates the point that product design and development is about tradeoffs. When engineers are concerned with optimizations of multiple variables, several tools exist to point us down the right path. Simulation gives us the chance to more efficiently find the ten or twenty possibilities that can give the best results with minimal disruption.

This is a great story with several engineering wins. TRINA researchers are using simulation in different ways than traditional engineers might approach a design project. At its core, simulation is used to give superior and faster results to these huge problems that come from heavy applications of math. Researchers and scientists might understand the gradients, three-dimensional vector valued functions and transformation functions on a different level than simulation engineers but the reality is that using simulation tools can solve these systems of equations much faster than a human can. Sometimes it makes sense to find the best possibilities for designs before running computation-heavy studies, and some faster studies can run dozens or hundreds of parallel paths to find the best possible combination of variables for optimal results.

This application of COMSOL from the TRINA researchers is a new design approach that shows how simulation engineers can use simulation for non-traditional problems. Interesting twists on this approach are the elements of biomimicry that come in when Zhou talked about the team’s attempts to find an idealized topology for flow vents. Breaking down the output-based problem to its core, the team then transferred these ideas into thermal, structural and acoustic applications for another win.

If simulation engineers and researchers are going to solve the big picture problems in the next one hundred years, several incremental solutions will be required to create large-scale improvements. It will be interesting to watch Toyota’s Woven City project and see how hydrogen is able to help the world become more sustainable and carbon neutral. Even more interesting will be watching the ways that simulation can help to make hydrogen a more viable energy source.