Secure Online Digital Twin Incorporating Physics-Aware Machine Learning for Additive Manufacturing

Currently, there isn’t a digital twin framework that allows for real-time interactive performance monitoring and control of manufacturing systems. A big challenge in developing such a digital twin is accelerating computations so they can be done in real-time. (Machine-learning algorithms, with their generalizing capability, are an ideal substitute for compute-intensive manufacturing simulations.) Additionally, real-time control and interconnectivity of machine tools increase the threat of cyberattack, so cybersecurity tools must be incorporated into this framework.

The team will develop a real-time, online digital twin to enable closed-loop control of advanced additive manufacturing processes. It will leverage physics-aware machine learning to connect sensor and process parameter output to material properties; high-speed graphics-processing unit (GPU) computing to enable real-time control; and cybersecurity principles to support trust and provenance in manufacturing line security.

With MxD funding, the team will create and deploy a novel digital twin framework for additive manufacturing processes, leading to efficient fabrication of high-value components. This solution will for the first time use a digital twin and machine learning for real-time control in manufacturing and will mitigate cybersecurity threats by incorporating cybersecurity measures into the framework. Ultimately, collaboration with industries will enable wide adoption of the methodology by manufacturers.