Determination of Dimensional Design Parameters of Pyramidal Lattice Core Sandwiches by Artificial Neural Network Under Bending Load

Abstract

This thesis investigates the effect of dimensional design parameters on the bending behavior of pyramidal lattice truss core sandwiches using artificial neural networks (ANNs). Pyramidal lattice truss core sandwiches are advanced structural materials known for their high specific strength and stiffness, making them ideal for aerospace and mechanical engineering applications. This study aims to develop an ANN to predict the mechanical response of these structures under bending loads, thereby reducing the time required for component analysis. The research begins with the fabrication of pyramidal lattice truss core sandwiches using additive manufacturing techniques, specifically fused deposition modeling (FDM). The cores are made from polylactic acid (PLA) reinforced with milled carbon fibers. The deformation behavior of the fabricated structures is experimentally evaluated through four-point bending tests, conducted according to ASTM D7250 standards. A linear static finite element (FE) model is developed to simulate the bending tests and validated against experimental results. The validated FE model is used to generate a comprehensive database, reflecting the influence of different geometric parameters on structural behavior. An ANN is then trained using this database to predict the mechanical response of the pyramidal lattice truss core sandwiches. The ANN algorithm takes dimensional design parameters—such as sandwich height, rod diameter, facesheet thickness, inclination angle, and the number of unit cells—as inputs and provides reaction forces and maximum stress components as outputs. The integration of machine learning enables rapid and accurate predictions, significantly enhancing the design and optimization process. The results demonstrate that the ANN can effectively predict the mechanical performance of pyramidal lattice truss core sandwiches under bending loads, offering a substantial reduction in the time required for design and analysis. This research contributes to the broader understanding of the interplay between material properties, geometric design, and mechanical performance, facilitating the optimal design and application of these innovative composite structures in engineering contexts.