Defect identification of wind turbine blades based on defect semantic features with transfer feature extractor

The monitoring of the status of the wind turbine blades is significant for the wind generation system and currently mainly dependent on manual visual inspections. The variance of the blade defects and the lack of the blade defect images make the defect identification of the wind turbine blades challenging. This paper proposes a defect identification method of wind turbine blades based on defect semantic features with transfer feature extractor. A deep convolutional neural network (DCNN) is built and is trained on the ImageNet Large Scale Visual Recognition Challenge dataset. The deep hierarchical features of the training blade images are extracted by the trained DCNN and fed into a classifier. By training on the labeled blade images, the first n layers of the trained DCNN is selected as the transfer feature extractor to extract the defect semantic features and the defect classifier is also obtained. The blade images can be diagnosed by the defect classifier based on the defect semantic features. The experiments are conducted on a real dataset of wind turbine blade images. The experimental results demonstrate the high learning ability of the proposed method from the small samples and its effectiveness for the defect identification of wind turbine blades. (C) 2019 Elsevier B.V. All rights reserved.