Research On Defect Type Identification Of Wind Turbine Blade Based On Infrared Detection Technology

With the state’s proposal that carbon dioxide emissions should reach the peak before 2030 and strive to achieve the goal of carbon neutralization by 2060,wind power generation as a clean energy has attracted extensive attention,and the cumulative installed capacity of wind turbines is increasing year by year.As the main key component,the wind turbine blade has frequent faults due to its long-term exposure to the complex and extremely harsh environment,which will directly affect the safe operation state of the wind turbine.And in the statistical results of wind turbine fault data,the shutdown fault of wind turbine caused by blade damage accounts for a very high proportion.Therefore,it is of great significance to conduct regular inspection and safety maintenance of wind turbine blades.The commonly used nondestructive testing methods include ultrasonic testing,acoustic emission testing,radiographic testing and infrared testing technology.However,ultrasonic testing and acoustic emission testing require strict experimental conditions,and the detection area is small,which is not suitable for large-area wind turbine blade detection.Radiographic testing is expensive and has some limitations in blade testing.Therefore,in the process of wind turbine blade detection,it is necessary to improve the detection ability of infrared detection technology.This paper takes the wind turbine blade as the research object,aiming at the problem of low definition and serious noise interference of the blade image collected by the infrared thermal image recorder,which makes it difficult to identify the defect and fault.In this paper,the infrared thermal wave detection system is built,the heat transfer principle of heat conduction theory in different substances is deduced,and the research work on the infrared image processing and blade defect classification of wind turbine blades is carried out by using the research methods of mathematical morphology and deep learning theory,combined with experimental testing methods.The main contents of the paper are as follows:(1)Based on the heat conduction theory,the heat conduction mathematical model of different substances in the heating process is established,the phenomenon of the maximum temperature difference of different substances in the continuous heating process is obtained,and the root cause of the change of infrared image from fuzzy to clear to fuzzy in the continuous heating process is revealed.Then,the manufacturing process of wind turbine blade is simulated,and the defect experimental samples are prepared,the effectiveness of the mathematical model of heat conduction is verified by simulation and experiment.By setting an appropriate temperature threshold,the preliminary judgment of prefabricated defects in the test sample is completed.Make preparations for the early stage of infrared detection of wind turbine blades for high-altitude operation.(2)Aiming at the problems of noise interference and low edge definition of the collected infrared image,the noise reduction effect of wavelet transform,mean filter and median filter on the gray image is studied.Then from the perspective of edge detection,combined with morphological gradient operator and first-order differential theory,a differential basic morphological gradient detection operator(DBMG)is proposed to extract the edge feature information in the image,Using the contrast value based on the principle of four adjacent as the evaluation index,the contrast values of Sobel,Prewitt,Dilate-erode and DBMG operators are compared.The results show that DBMG operator is obviously better than other operators.On the basis of this research,the image fusion technology is used to weighted fuse the obtained edge information with the original image,and the contrast limited adaptive histogram equalization(CLAHE)is used to further improve the overall effect of the image.The experimental results show that the edge definition of the image obtained by this method is high,and the effect of practical engineering application is good.(3)In order to improve the detailed texture features of infrared images,starting from the structural elements of multi-scale and differential scale,a differential multi-scale morphological top hat transformation enhancement algorithm is proposed.The image stability is improved by changing the scale value and differential scale value,and then the problem of poor stability of traditional morphological top hat operators is overcome.In order to solve the selection problem of scale value and difference scale value,the contrast improvement factor is introduced as the index of scale selection,and an adaptive threshold iterative weighted interval selection method is proposed.The optimal scale interval is selected through threshold iteration,and finally the details of the image are effectively enhanced.In order to verify whether the defect size in the infrared image is consistent with the actual size,grind the blade with defect area,and measure the defect size after grinding.The measurement results show that under the condition of ignoring the subjective measurement error,the error between the defect size in the collected infrared image and the defect size after actual grinding is less than 6%,which proves the accuracy of the infrared thermal wave detection system in quantitatively measuring the deep defect size of the blade.the accuracy of the infrared thermal wave detection system in quantitatively measuring the deep defect size of the blade.(4)Aiming at the problem that the existing algorithms extract less image feature information of blade defects and insufficient infrared image data samples of blade defects,this paper establishes a small sample blade defect database,and proposes a classification algorithm of multi-scale convolutional neural networks(MSCNN-SE)with compression excitation module,which adopts the parallel feature extraction method of multi-scale convolution kernel,The channel attention mechanism is used to optimize the fused results,so as to extract more abundant information features.The experimental results show that the proposed method has better classification accuracy and robustness in blade defect classification and detection.Finally,in order to realize the detection task of wind turbine blade under harsh conditions,and complete the spatial positioning and size measurement of defect position in the blade.In this paper,the infrared thermal wave detection system is built,and the infrared thermal imager is configured to develop a set of detection software for image acquisition,processing and classification.The image processing algorithm and trained classification model are embedded in Gideon infrared software to form a complete set of wind turbine blade infrared thermal wave detection system.