Session: 01-05-01: Foundations of AI and Machine Learning for Aerospace Structural Analysis

Paper Number: 183233

183233 - Multiscale Modeling of Composites With AI/ML Framework Using JAX 

Accurate modeling of composites requires high computational cost and resources. The Mechanics of Structure Genome (MSG) brings a revolutionary and unified approach for performing macroscale structural analysis with microscale information with the possibility to avoid scale separation assumptions and assumptions within each scale. MSG minimizes the loss of information between the original heterogeneous structure and the desired macroscopic model. The industrially trusted commercial tool for MSG implementation, SwiftComp, provides the beam, plate/shell, or Cauchy continuum model by performing homogenization and dehomogenization. MSG is based on the concept of structure gene (SG), which is defined as the smallest mathematical building block of the structure, capturing all the necessary heterogeneity and anisotropy, to perform multiscale constitutive modeling of the structure. The present work implements MSG to create an accurate and efficient multiscale modeling tool, called OpenMSG, in JAX for predicting mechanical and multiphysics behaviors of heterogeneous structures and materials. The parallel computing of GPU/TPUs integrated with AI/ML addresses long-standing issues of efficient modeling of multiscale problems. Automated differentiation, functional vectorization, and just-in-time (JIT) compilation scale the sequence of operations in multi-GPU/TPU clusters through XLA. This advanced XLA approach doesn’t require any external directives like OpenACC to initiate the GPU accelerators for only specific loops. The robust finite element implementation in JAX, featuring one, two, or three-dimensional SG with desired functional space, will address the complex multiscale behavior of composite structures and materials. The MSG implementation will not only use solid elements but will also have extended capability to use lattice microstructures and shell geometries for predicting macroscale behavior. We will perform extensive validation of MSG implemented solid models with SwiftComp to validate OpenMSG’s accuracy and efficiency. OpenMSG enables a single simulation to leverage the AI/ML ecosystem for major aspects, including constitutive relations, local buckling, and progressive damage. OpenMSG will also be integrated with other modules of the fea-in-jax framework for multiscale simulations of composite structures involving multiscale scales. The proposed OpenMSG will provide an unprecedented potential for a hybrid of data-driven and physics-based multiscale simulation to address more complex innovations in composite design and manufacturing. OpenMSG will be publicly available in the fea-in-jax GitHub repository with API-based implementation where the research community can leverage MSG with an AI/ML integrated framework as a better alternative than finite element analysis of representative volume elements (RVEs).

Presenting Author: Akshat Bagla Purdue University

Presenting Author Biography: Akshat Bagla is a fourth year Ph.D. student in Aerospace Engineering at Purdue University in Prof. Wenbin
Yu’s research group. His research focuses on creating an AI/ML framework for multiscale modeling of composites, and local buckling of wind turbine blades. Akshat is working on NSF project for developing OpenMSG, an open-source general purpose multiscale code using JAX. Akshat received his master’s in aerospace
Engineering from IIT Kanpur, India and worked in R&D team of structural design of mini helicopters in EndureAir
Systems Pvt. Ltd, India.