High-fidelity surrogate-driven h-refined IGA for free vibration analysis of laminated composite annular plates with radial and curved cracks

During their service life, structures often develop cracks, which affect their free vibration behavior. This study focuses on predicting the free vibration response of cracked laminated composite annular plates with helicoidal layup schemes, considering both radial and circular cracks. The plate is modeled using isogeometric analysis based on Reddy’s shear deformation theory. To accurately represent the crack, an adaptive h-refinement strategy is employed within the crack region, with the refinement level determined by a Gaussian process regression (GPR) surrogate machine learning model. The surrogate model is trained using preliminary free vibration analyses under varying boundary conditions, crack lengths, layup schemes, and plate geometries. Results indicate that second- level refinement is sufficient for modeling radial cracks, whereas higher refinement is necessary for complex circular cracks. The present data-driven approach enhances computational efficiency while ensuring accurate vibration analysis of cracked composite plates.