Effectiveness of Artificial Neural Networks and Surrogate-Assisted Optimization Techniques in Delamination Detection for Structural Health Monitoring

Obinna K. Ihesiulor, Krishna Shankar, Zhifang Zhang, and Tapabrata Ray


Composite laminates, Vibrations, Finite element methods, Delamination detection, Artificial neural networks, Optimization techniques


This paper focuses on the viability of integrating K-means clustering, artificial neural network (ANN) and optimization techniques (Non-dominated Sorting Genetic Algorithm-II (NSGA-II) and gradient based local search) to effectively detect delamination parameters (interlaminar position, location and size) in composite structures for efficient structural health monitoring (SHM). Finite element (FE) method is harnessed to determine natural frequencies of a laminated composite beam with and without delaminations. The high-fidelity numerical simulations are computationally expensive and therefore, their direct use in the optimization loop is inhibitive. This necessitates the need to develop computationally efficient and cheaper design via artificial neural network (ANN) surrogate-based modeling for enhancement of the optimization process. ANN is also employed to detect accurately delamination parameters using Bayesian regularization, a robust iterative training algorithm. Prediction results are shown to be efficient in terms of accuracy.

Important Links:

Go Back