Session: 01-05-02: AI-Driven Modeling and Simulation for Aerospace Structures

Paper Number: 190382

190382 - Understanding the Structure-Property Relationship of I-Beam Lattices Using Gaussian Process Regression and Explainable Ai 

Beam lattice metamaterials have demonstrated outstanding stiffness- and strength-to-weight ratios as well as superior energy-absorption capabilities, making them increasingly attractive for aerospace applications in recent years. Depending on their dominant deformation mechanisms, beam lattices can generally be classified as stretching-dominated or bending-dominated. Bending-dominated lattices rely primarily on the flexural response of their beams, with failure typically governed by joint yielding. This deformation mode enables structures such as body-centered cubic (BCC) lattices to sustain extended stress plateaus and absorb large amounts of energy, rendering them promising candidates for impact-protection systems.

Despite the wide range of lattice topologies explored in the literature, BCC lattices continue to receive significant attention due to their stable crushing response and excellent energy dissipation performance. Recently, the authors proposed reinforced I-beam BCC lattices with joint reinforcements, which exhibit substantially enhanced stiffness, strength, and energy absorption compared with conventional designs. These improvements are primarily attributed to increased bending stiffness of the beam cross-sections and a novel two-step energy dissipation mechanism. However, a critical challenge in advancing this new class of beam lattices lies in understanding the complex structure-property (SP) relationships, as the crushing response is highly sensitive to beam cross-sectional geometry and reinforcement design. Previous studies have not systematically investigated the influence of these geometric parameters on key mechanical properties (e.g., stiffness, strength, and energy absorption), thereby limiting the full exploitation of reinforced I-beam (R-I-beam) lattices.

Addressing this challenge requires a comprehensive exploration of the design space to quantify how sectional geometric parameters and their interactions affect lattice crushing behavior. Experimental investigations are often impractical due to high cost and extensive testing efforts. As an alternative, finite element (FE) simulations have been widely used to study SP relationships in beam lattices; however, accurately capturing the highly nonlinear deformation and failure processes remains computationally expensive. Recently, machine learning (ML) techniques have emerged as efficient surrogate modeling approaches. In particular, Gaussian Process Regression (GPR) models have demonstrated strong predictive capability for small-data applications.

In this work, crushing simulations of I-beam and reinforced I-beam lattices with varying sectional design parameters are performed. Mechanical properties, including stiffness, yield strength, and energy absorption, are extracted from the simulated stress-strain responses. GPR models are then developed to predict these properties using sectional geometric parameters as inputs. To further interpret the learned relationships, the GPR models are coupled with explainable artificial intelligence (XAI) techniques. The results demonstrate that the proposed framework achieves high predictive accuracy and reveals that sectional geometries play a dominant role in governing the energy-absorption performance.

Presenting Author: Xin Liu The University of Texas at Arlington

Presenting Author Biography: Dr. Xin Liu is an Assistant Professor in the Department of Mechanical and Aerospace Engineering at the University of Texas at Arlington and a faculty member of the Institute for Predictive Performance Methodologies (IPPM) at the UTA Research Institute. He earned his Ph.D. (2020) and M.Eng. (2016) in Aeronautics and Astronautics Engineering from Purdue University. Dr. Liu’s research focuses on machine learning-assisted multiscale modeling of materials and structures.