Wind Turbine Blade Fault Detection Based On Deep Learning

As one of the key components of wind turbines,blades are critical to the efficiency and safety of wind power generation.However,traditional wind power fault detection is mostly based on the processing of sensor signals,which has low efficiency and poor robustness.The development of deep learning in recent years has provided a novel idea for the question.This article focuses on this,and conducts in-depth research and implementation on how to implement defect detection based on deep learning.The main work is as follows:(1)Investigate the relevant background of the research content,study the application of deep learning technology in the detection of fan blade defects,and do research on how to distinguish different defects,and design algorithm models based on actual application scenarios.(2)An improved Mask-RCNN WT blade fault detection method is proposed.The algorithm uses Res Net-50 to combine the FPN network to generate feature maps,and then input these into the RPN network to filter out the ROI.Finally,the proposed algorithm fixed the dimension of ROI by ROIAlign and input these into the prediction network for classification,object detection,and instance segmentation.In addition,this article builds a mask dataset of WT blade fault for model training.This algorithm achieves the object detection and instance segmentation of fault,and greatly improves the detection time at the same time.(3)Proposed Mask MRNet for wind turbine blade fault detection.In the actual application process,it was found that most of the image of the blades taken was inclined,which easily caused the accuracy of the model to drop.Mask MRNet was designed to correct the defects of the fan,which greatly improved the classification effect of the defects,and the accuracy rate reached 95%.(4)Designed a wind turbine blade fault detection system and deployed the model to the backstage.The system can analyze the blade pictures or video data,and can intuitively push the recognition results to the user interface,which greatly facilitates the daily inspection of the staff.