Damage Detection Based on Self-Organizing Map Neural Network

Structure damage is a great threat to an uninterrupted operation of modern machines because it may cause catastrophic failures. Thus, damage detection has become the most important research topics. At present, a large number of damage detection methods have been proposed and applied to the field of structural damage detection, among which the most widely used detection methods are based on vibration analysis. On this basis, we proposed a method of combining short-time Fourier transform (STFT) and pulse-coupled neural network (PCNN) to extract signal characteristic, then use the signal to train self-organizing map (SOM) neural network to classify and identify of structural damage.