Fault identification and classification of wind turbine blades based on improved DenseNet

As the core part of the wind turbine, the fan blade often fails, which has great potential safety hazards. In order to detect the faults of wind turbine blades timely and effectively, an improved DenseNet network is proposed to identify and classify the faults of wind turbine blades. The Xception multi-layer convolutional structure is integrated into the DenseNet network. The dense block structure is enhanced by incorporating the Xception module into the bottleneck layer, and the spatial and channel characteristics are separated using the Xception depthwise separable convolution structure Combined with the residual structure of DenseNet network dense connection, the classification and identification are carried out. Through experimental comparison, it is found that the improvement of DenseNet model in this paper has higher classification accuracy than before, and effectively saves the classification time.