Implementation of neural networks for prediction of location and orientation of pipe defects in guided wave ultrasonic testing

In this research, a method based on the the artificial neural network is used to determine the location and orientation of cracks in pipes. For this purpose, first, the finite element method is used to model wave propagation and crack modeling in different locations and orientations. In this regard, two types of longitudinal and torsional guided waves are used to excite the structure. The obtained signals are processed in order to calculate the appropriate characteristics. To this end, the reflection echoes are also measured and five features are extracted at six levels from discrete wavelet decomposition of raw signals. Selected properties of the signals are processed to limit the size of the neural network model without losing information. To do so, the firefly algorithm method was used and fed to an artificial neural network that predicts the location and orientation of the crack. In this study, conventional multilayer perceptron diffusion networks have been used. According to the obtained results, it is observed that the proposed method shows good accuracy in predicting the location and orientation of the crack. Also, the percentage of neural network errors is less than 7%.