UC San Diego Engineers Lead $5M DOE Center to Design Impact-Resistant Metamaterials

This is in large part because creating a computational model to understand how these materials would behave and determine the optimum spacing and patterning of the internal geometries are considered impossible. The Center will develop a computational topology optimization method which will automatically design and discover new nonlinear materials via both numerical and experimental studies.

Such impact-resistant materials have direct applications in the defense and space sectors, but the ability to successfully design materials with specified nonlinear dynamic responses would also have far-reaching applications in robotics and mechanical computing.

This new Center, called SHAPE – Center for Simulation and design of Heterogeneous Architectures for Performance and Energy absorption – brings structural engineers, mechanical engineers and computer scientists across UC San Diego together with mechanical engineers from Colorado School of Mines to solve these computing and engineering challenges. The Center is part of the NNSA’s Predictive Science Academic Alliance Program (PSAAP).

"When high speed projectiles hit a material, sometimes damage is caused not by the impactor puncturing the material but by a shock wave that it induces in the material,” said Nicholas Boechler, associate professor of mechanical and aerospace engineering at UC San Diego and a Center investigator. “One of the goals of our Center is to use the idea that, by changing the way a material  behaves nonlinearly in response to stresses and by introducing spatial patterning, we can make materials that are more resilient and better protective devices.”

In order to make this possible, the multidisciplinary research team is exploiting new Accelerated Processing Unit (APU) computing architectures; novel means of computational modeling and optimization; and applying structural engineering deformation modeling techniques to these nonlinear materials.

Handling the computational load

The computations required to understand how these new materials will behave under various forces and to determine exactly where and in what shape these internal geometries should be placed to achieve the desired nonlinear properties are so large that it’s not currently possible for existing tools to produce an optimized design. SHAPE will leverage the Tuolumne supercomputer at the Lawrence Livermore National Laboratory to overcome this computing challenge. Tuolumne, funded by NNSA, includes APUs which provide both CPU and GPU compute units that share a common memory pool. This flexibility allows for gradual GPU-acceleration of algorithms, overcoming the paradigm shift chasm that doomed previous GPU-only endeavours. Paired with the large scale of the system, which provides more than 4500 APUs in a fully-connected HPC setup, the simulation time of non-linear computational models becomes short enough to provide actionable insights.

While Tuolumne will eventually provide the bulk of computational resources, the researchers will start the initial development and prototyping activity on the new Cosmos supercomputer at the San Diego Supercomputer Center. The Cosmos system, funded by the National Science Foundation, provides a very similar APU-based computational HPC environment. Having an additional local resource will significantly improve the productivity of the team.

The researchers will also develop  a new physics-guided geometric multigrid optimization method to speed up the work of understanding the many nonlinear responses happening within the structure as it is manipulated.

Meshfree Models

Another challenge of designing materials with nonlinear properties is that the tools used to model materials with natural properties often don’t produce reliable results when dealing with nonlinearities. For example, the simulation models used to study deformation in existing structures often use a mesh-type grid to model how the structure bends and breaks. With internal geometries, this becomes much trickier. JS Chen, a Center investigator and professor of structural engineering at UC San Diego, pioneered a more accurate particle-based method to study shape change and deformation. The method, called RKPM, is now widely used for  various structures and materials. The SHAPE team will now apply it to nonlinear metamaterials and add neural networks and machine learning to speed up the process.

“We are excited to welcome the SHAPE Center to the fourth phase of the PSAAP program," said David Etim, PSAAP Federal Program Manager from the NNSA Office of Advanced Simulation and Computing and Institutional Research & Development. "Their research into impact-resistant metamaterials, leveraging advanced computational methods and high-performance computing, is critical to national security challenges within NNSA’s mission and promises to advance predictive science capabilities.”

SHAPE team: UC San Diego faculty Alicia Kim, professor of structural engineering; Boris Kramer, associate professor of mechanical and aerospace engineering; Nicholas Boechler, professor of mechanical and aerospace engineering; JS Chen, professor of structural engineering; Igor Sfiligoi, senior research scientist at the San Diego Supercomputer Center; and Leslie Lamberson, associate professor of mechanical engineering at Colorado School of Mines