A TWO-STAGE METHOD FOR DETECTING DIAGONAL CRACKS BASED ON GENETIC NEURAL NETWORK AND MODAL STRAIN ENERGY

Based on genetic neural network and modal strain energy, a two-stage method for detecting diagonal cracks is proposed to identify the location, angle and depth of diagonal cracks in beams. According to linear elastic fracture mechanics and virtual work principle, the elemental stiffness matrix of a diagonally cracked beam is derived, and the frequencies and modes of the diagonally cracked beam are obtained. The topological structure, weight and threshold of the BP neural network are optimized using the genetic algorithm, and a genetic neural network is built to identify the location and angle of the diagonal cracks in beams. By combining the modal strain energy of the diagonally cracked element and integrating the stress intensity factor of the diagonally cracked element, the analytical expression of the depth of the diagonal crack is obtained to identify the depth of the diagonal cracks. The numerical simulation shows that the proposed method may detect the damage state, including the location, angle and depth, of the diagonal cracks in beams with high precision. By comparing with the BP neural network, the genetic neural network has stronger generalization capacity and better robustness against measuring noises.