An Improved Defect Detection Method of Water Walls Using the WGAN

This paper proposes an improved water wall defect detection method using Wasserstein generation adversarial network (WGAN). The method aims to improve the problems of poor safety and high level of maintenance personnel required by traditional inspection methods, and improve the accuracy and safety of water wall defect detection through automated inspection. The WGAN is used to expand the water wall defect data. Then the extended data set and the original data set are loaded into a convolutional neural network for training, respectively. The results show that the accuracy of the expanded data set is significantly improved, which can achieve the industrial requirement of the thermal power generation.