Research on Detection of Surface Defects of Wind Turbine Blades Based on Deep Learning
The detection of wind turbine blade damage mainly relies on visual inspection and percussion,not only inefficient,but also easily affected by human subjective judgment factors.To solve these problems,this paper proposes an image recognition method based on ResNet50 convolutional neural network based on Rectified Adam(RAdam)optimizer,classification and recogni-tion of wind turbine blade damage images.Drones is used to take pictures of the damage location of wind turbine blades,the collect-ed images are filtered and enhanced to obtain a data set of four types of leaf damage.The image is subjected to undergoes grayscale processing,denoising,and threshold segmentation to remove the influence of the background information of the image.The recogni-tion accuracy of the three network models of VGG19,GoogleNet,and ResNet50 are analyzed for the type of wind turbine blade dam-age.The ResNet50 network model with higher classification accuracy is selected.Compared and experimented the accuracy of ResNet50 for wind turbine blade damage recognition under the two optimizers of Adam and RAdam,the results show that the perfor-mance of the ResNet50 network model under the RAdam optimizer is better,a reference is provided for the automation and digitiza-tion of the non-destructive testing of wind turbine blades.
国家科技期刊平台
NSTL
万方数据
