Research On Wind Turbine Blade Damage Detection Method Based On Image Recognition Technology

The share of wind power in energy is constantly increasing.Blade is the main component of wind turbine,and the cost is high.In use,it is affected by material degradation,fatigue,frost,lightning and other factors,resulting in blade damage failure,which has a negative impact on the operation of the wind turbine.Regular inspection of wind turbine blades can effectively reduce the occurrence of faults and accidents by finding potential damage and timely maintenance.Therefore,the damage detection of wind turbine blade is studied.Firstly,the image set of wind turbine blades is expanded.Due to the small scale of the existing data set,the occurrence of some damage types and the number of images in the image set,it is necessary to expand the data set in order to enhance the reliability of the system model and improve its generalization performance.Based on the algorithms of perspective transformation,flipping,contrast normalization and Gaussian blur,the image set is augmented.At the same time,in order to standardize the size,pyramid scaling and splicing method are used to adjust.The target image is preprocessed to highlight the characteristics of blade damage.Secondly,for non-destructive defect detection,highlight the edges of damage.When designing the system,a multi-scale edge detection algorithm is used to identify the edge of the wind turbine blade image,compare the value of each pixel with the pixels in the adjacent area,and use 3×3 two-dimensional discrete convolution to calculate,thereby improving the system Recognition accuracy and generalization performance.Thirdly,in order to detect defects in wind turbine blades,deep learning algorithms such as BP neural network,CNN network,and GoogLeNet are studied.Among them,GoogLeNet has good sparse characteristics,which can greatly reduce the network depth and reduce the computing power scale.The system adopts the Tensor Flow framework.In order to adapt to the application of the system and reduce the occupation of computing resources,the Inception architecture is improved,and the improved GoogLeNet algorithm is established.The blade surface defect detection is realized,and a satisfactory detection model is obtained.The detection accuracy is stable above 95%.Finally,the system is applied to Jiangsu Guoxin wind power project,and the terminal application interface is developed to facilitate the deployment and application of the wind turbine blade damage detection system.The interface consists of display,result record,menu toolbar and other functional areas.The feasibility of GUI interaction and the effectiveness of the system are verified by actual operation.