Research On Intelligent Image Recognition Technology For Wind Turbine Blade Damage

Wind energy is one of the most important renewable energy in modern society,and its main form of utilization is wind power generation.Wind turbine blades are the main components that capture wind energy and convert it into electricity.Due to the complex and harsh working environment of most wind turbines and the long-term bearing of alternating load,the blades of in-service wind turbines are prone to failure,thereby reducing the efficiency of wind turbine power generation and causing hidden dangers to safe production.Therefore,it is particularly important to detect and identify blade surface faults.Aiming at the problems in blade fault diagnosis,this paper analyzes the characteristics of existing diagnostic methods,combines the new technology of image recognition based on deep learning and the new trend of inspecting blades by UAV,and proposes a new idea for intelligent image recognition of blades defects based on deep learning.In this paper,the deep learning image recognition technology is studied deeply in the recognition of blade surface defects,and the intelligent image recognition models of blade surface defects are developed.The feasibility of the models is verified by case simulation,which can better solve the safety,efficiency,cost,reliability and other problems in the detection of blades surface defects.The main contents of this paper are as follows:1.This paper introduces the background and significance of the research,reviews the research status of blade fault diagnosis,summarizes the characteristics of various diagnosis methods,and analyzes the advantages of image recognition technology based on deep learning algorithm in blade defect recognition,combining with the increasingly mature image recognition technology and the emerging trend of using UAV to inspect blades.2.This paper studies the blade image,analyzes the difficulties in blade defect recognition,and uses the Otsu threshold segmentation method to preprocess the blade image to reduce the interference by the blade environment background to the defect recognition.Drawing on the extensive success cases of applying deep learning image recognition technology,applying the AlexNet model based on convolutional neural networks to blades surface defect recognition.The proposed AlexNet-based intelligent image recognition model for blades defects is compared with the traditional machine learning algorithm SVM,and the experimental results verify the superiority of the proposed intelligent image recognition model.3.Focusing on the problems of insufficient training data and low sensitivity of AlexNet in intelligent image recognition of blades defects,an improved AlexNet intelligent image recognition model of blades defects combined with transfer learning and random forest are proposed,and the ability of abstract feature extraction is further improved by transfer learning,and the ability of blades image feature classification is improved by using the integrated learning method of random forest;the advantages of improved intelligent image recognition model of blades defects are verified through experimental comparison.4.On the basis of the above research,the identification of blade defects is further studied,and the surface defects of blades are classified and located.The target detection network in deep learning is introduced into the identification of blade surface defects,an intelligent blade defect recognition model based on Mask-rcnn was proposed,which improved the ability to identify subtle defects of blades through RPN network,FPN structure,ROIAlign operation,and multi-tasking Loss,and realized the identification of blade defect types,defect location,and defect contour description.