Hierarchical Neural Network and Simulation Based Structural Defect Identification and Classification

A vibration data-driven structural defect identification and classification technique is developed using frequency response under random excitation a hierarchical neural network. system of artificial networks (ANNs) trained finite element simulation-based synthetic data to reduce the need for many sensor measurements required otherwise. Principal component analysis (PCA) employed compress high dimensionality eliminate noise effect in training testing. Frequency responses dimension structure with defects such crack from stress concentration, rivet hole expansion, attached foreign object mass as ice accumulation aircraft wing or fuselage are reduced PCA fed classifier The probabilistic decision output network compressed then next levels estimator networks, where each dedicated individual type estimation parameters corresponding that class defect. methodology applied stiffened panel structure. cracks expansions introduced line stiffener, surface. results show it possible classify further estimate good accuracy reliability. It was observed damage had an roughly 95%. localization well expansion average absolute error around 2. severity also able perform mean about 0.34 length detection 0.22 expanded damage. However, prediction were quite challenging train presence multiple damages development architecture.