Session: 13-06-01: Multiscale Models and Experimental Techniques for Composite Materials and Structures I

Paper Number: 173660

Constraint-Aware Topology Optimization of Additively Manufactured Composites 

In an era where sustainable and high-performance engineering solutions are critically needed, this work introduces a transformative framework that redefines structural design through the synergistic integration of topology optimization (TO), composite additive manufacturing (AM), and artificial intelligence (AI). This research responds to the urgent demands of aerospace, automotive, biomedical, renewable energy, and infrastructure industries for lightweight, durable, and eco-efficient structural components. By merging advanced computational design with cutting-edge manufacturing technologies, the framework sets a new standard for achieving optimized performance, manufacturability, and environmental sustainability, establishing a compelling paradigm for innovation in modern engineering.

At the heart of this work is a novel topology optimization methodology tailored for fiber-reinforced polymer composites (FRPCs), which are increasingly used in high-performance applications due to their exceptional strength-to-weight ratios and customizable mechanical properties. Unlike conventional TO methods, which often neglect material anisotropy and fabrication constraints, the proposed approach is designed specifically for advanced composite AM processes. Techniques such as continuous fiber-reinforced 3D printing and 3D Fiber Tethering (3DFIT) are leveraged to fabricate structurally optimized components while addressing critical manufacturing limitations, including discrete fiber orientations, support scaffold configurations, and build sequence constraints. This method allows for the simultaneous optimization of structural layout and fiber path, enabling the production of geometrically complex components that maintain mechanical integrity under demanding conditions.

A key demonstration of this methodology is its application to the design of an automotive B-pillar, a structural element subject to stringent crashworthiness and weight-reduction requirements. Through the co-optimization of geometry and fiber placement, the proposed method achieves significant weight savings without compromising safety performance. The resulting structure demonstrates superior crash energy absorption and improved material efficiency, highlighting the practical potential of the approach in real-world engineering challenges. Furthermore, the reduced material usage and improved structural performance contribute directly to sustainability goals, minimizing waste and enabling greener manufacturing pathways.

To extend the capabilities of the framework and address the high computational cost of traditional TO methods, an AI-driven design acceleration strategy has been introduced. A novel dual-stage ResUNet architecture is proposed, coupled with a physically informed hybrid loss function that enforces both topological and mechanical constraints. This hybrid loss function integrates distance-based and similarity-based components to promote structural connectivity, reduce compliance error, and improve the visual and functional quality of generated topologies. The AI model significantly reduces computation time—achieving near-optimal solutions up to 400 times faster than conventional solvers—without compromising design quality. The model’s predictive power is further enhanced through an iterative refinement stage that elevates low-resolution predictions to high-fidelity, manufacturable designs, thereby enabling rapid design exploration and decision-making across multidisciplinary applications.

This integration of AI and composite AM into the TO workflow represents a major advancement in structural optimization, overcoming longstanding barriers of computational cost, manufacturability, and design iteration speed. The framework offers not just incremental improvements, but a reimagining of the design process itself—one that is intelligent, adaptive, and deeply aligned with the requirements of sustainable development. By embedding physical principles into data-driven algorithms and tailoring optimization strategies to advanced manufacturing capabilities, this work enables the creation of next-generation structures that are lighter, stronger, and more environmentally responsible.

Presenting Author: Ling Liu Temple University

Presenting Author Biography: Associate Professor of Temple University

Authors:

Md Mohaiminul Islam Temple University
Ling Liu Temple University