Damage Localization in Pressure Vessel by Guided Waves Based on Convolution Neural Network Approach

Abstract

This paper investigates the damage localization in a pressure vessel using guided wave-based structural health monitoring (SHM) technology. An online SHM system was developed to automatically select the guided wave propagating path and collect the generated signals during the monitoring process. Deep learning approach was employed to train the convolutional neural network (CNN) model by the guided wave datasets. Two piezo-electric ceramic transducers (PZT) arrays were designed to verify the anti-interference ability and robustness of the CNN model. Results indicate that the CNN model with seven convolution layers, three pooling layers, one fully connected layer, and one Softmax layer could locate the damage with 100% accuracy rate without overfitting. This method has good anti-interference ability in vibration or PZTs failure condition, and the anti-interference ability increases with increasing of PZT numbers. The trained CNN model can locate damage with high accuracy, and it has great potential to be applied in damage localization of pressure vessels.