| This study examines the curing behavior of composite laminates by combining finite element simulations with data-driven surrogate modeling. A cure kinetics model is implemented in ANSYS Composite Cure Simulation to generate a large parametric set of cure histories for representative laminate configurations. The design space covers a wide range of processing, material, and structural parameters, including heating ramps, dwell temperatures and times, laminate thickness, number of plies, total heat of reaction, symmetric and unsymmetric stacking sequences, and intentionally non-optimal cure cycles. In total, 1016 three-dimensional simulations are carried out. For each cure cycle, time-dependent histories of temperature, degree of cure, and maximum principal stress are extracted and collected into a comprehensive dataset. The input space is represented using compact descriptors of the cure cycle, laminate geometry, material properties, and layup statistics, while the model outputs the corresponding thermal, chemical, and stress evolution histories. A shared-task neural network surrogate is trained to learn the mapping from processing conditions and laminate characteristics to the cure response. To better capture rapid thermal transitions and exothermic behavior, the simulation histories are represented on a shared warped time grid with increased resolution in highly nonlinear regions. The trained model reproduces finite element predictions of temperature, degree of cure, and maximum principal stress across the explored design space with low computational cost. This forward surrogate provides a fast prediction tool for comparing cure-cycle responses, identifying conditions associated with incomplete cure or large thermal gradients, and supporting future physics-informed process design studies. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.