Machine learning-accelerated predictions of design allowable of composite laminates

The generation of design allowables for composite laminates is crucial for aerospace composite structure design and certification. Typically, determining these design allowables involves costly and time-intensive experimental tests. With advances in computational power and highfidelity numerical models that accurately simulate composite materials' responses, finite element simulations have emerged as alternatives to reduce certification costs, though they remain computationally demanding. Recent advancements in machine learning offer new possibilities for rapidly predicting materials' structural responses via surrogate models that describe the design space analytically and continuously. A database of open-hole high-fidelity simulations was used to train machine learning algorithms, creating surrogate models capable of predicting the notched strength of various materials, layups, and notched geometries. These trained algorithms can predict the strength of open-hole multidirectional composite coupons with high-fidelity simulation precision in milliseconds, achieving a prediction speed-up of over 10,000 times.