A hybrid metaheuristic framework for precision delamination detection in composite laminates

In this study, a two-stage hybrid metaheuristic framework is developed for the quantitative assessment of delamination damage in composite laminates using vibration responses. Layer-wise theory is employed to model composite structures in both undamaged and delaminated states, considering various damage scenarios with single- and multiple-interface delaminations. The framework utilizes the first few natural frequencies and mode shapes to identify and refine damage characteristics. In the first stage, a damage index based on mode shape curvatures approximates the in-plane location of delaminations in beams and plates. In the second stage, an inverse algorithm refines these estimates to pinpoint the precise location, size, and interfaces by minimizing an objective function derived from natural frequencies and mode shapes. The key novelty of this work is the innovative binary-discrete-continuous approach, which reduces the optimization search space by converting the binary representation of delaminated interfaces into a unique integer, thereby unifying the treatment of discrete and continuous variables and significantly enhancing convergence efficiency. The framework’s versatility is demonstrated in numerical simulations and experiments on glass-fiber-reinforced polymer beams, where it reliably identifies delaminations even under complex conditions. For composite plates, the approach proves effective under both noise-free and noisy environments, particularly for larger delaminations. The method is characterized by its simplicity, adaptability, and computational efficiency, requiring fewer control variables and maintaining compatibility with a range of optimization algorithms. These features, combined with its robust performance in both simulated and experimental settings, establish the proposed framework as a promising tool for delamination detection in composite structures with significant potential for structural health monitoring applications.