ADDITIVE MANUFACTURING AND DESIGN OF MECHANICALLY TAILOR-MADE HYBRID COMPOSITES WITH ARTIFICIAL INTELLIGENCE

Abstract

Hybrid composites are composed of multiple fillers incorporated into a single matrix. With enhanced and unique material properties, hybrid composites are suited for specific applications in material science and mechanical engineering, particularly in aerospace, automotive, and sports equipment. Hybrid composites with various fillers of different geometries and properties are multifunctional with complex microstructures that affect mechanical performance. Designing materials with specific characteristics is challenging due to the numerous microstructural factors, such as filler geometries and distributions, as well as the complexity of analyzing these structures. Traditional design methods rely on trial and error which are time-consuming and computationally intensive requiring extensive human intervention for feature extraction and descriptors selection. Recent advancements in materials science leverage artificial intelligence (AI) including deep learning (DL) to tailor microstructure designs for desired mechanical traits and discovery of novel materials through large-scale databases. The main goal of this dissertation is to develop an AI-driven framework for designing tailor-made hybrid composite microstructures with desired mechanical behaviors. By integrating AI with finite element analysis (FEA), additive manufacturing (AM), and experimental validation, the framework aims to establish a comprehensive understanding of structure-property (S-P) and inverse S-P relationships in hybrid composites. Firstly, the study aimed to characterize S-P relationships by leveraging a conventional design of experiments, a theoretical hybrid model, and an image-driven machine learning (ML) approach to investigate the mechanical behaviors of 3D printed lightweight hybrid composites. Then, inverse microstructural designs for hybrid composite materials were developed to achieve tailor-made full-range stress-strain relationships by utilizing FEA and DL techniques. Lastly, inverse S-P relationships were formulated by integrating a domain-specific constraint model which explained the design of tailor-made materials using fundamental knowledge of material design. Custom-designed lightweight materials with AI-driven models will enhance the understanding of S-P relationships in hybrid composites leading to reduced design time as well as computational resources.