V. Perumal
UHV3D, Inc. (dba CAMINNO), New Mexico, United States
Keywords: digital engineering, machine learning, design and manufacturing optimization, generative AI, hypersonics
Over the past five decades, design and manufacturing have relied heavily on subject matter experts, but current digital tools face limitations when tackling the complexity of advanced applications such as nuclear fusion and carbon sequestration. Traditionally, design and manufacturing challenges are addressed separately, constrained by both digital capabilities and human cognition. Advanced manufacturing (AM), including additive manufacturing, remains underutilized in high-tech and safety-critical industries due to quality, repeatability, and reliability issues. A key challenge lies in the stochastic nature of AM, where factors such as thermal history, flaw formation, microstructure evolution, and machine variability lead to inconsistent part performance. These uncertainties hinder manufacturing on-demand of customized components for aerospace, defense, and fusion technologies. The proposed solution, CAMINNO Gen6, introduces a multi-objective design optimization algorithm that integrates scientific machine learning (ML) with physics-based modeling and experimental data. This approach builds a manufacturing digital twin capable of predicting outcomes in near-real time. Surrogate models accelerate computation by analyzing process data, evaluating quality metrics, predicting thermal and mechanical behavior, and recommending design adjustments to ensure part performance. By enabling faster, more accurate design-manufacturing integration, digital twins can reduce costs, accelerate innovation, improve reliability, and lower overall carbon footprint of engineering and manufacturing.