Research On Electromagnetic Ultrasonic Detection Technology For Wind Turbine Blade Defects

The demand for green and clean energy is increasing in modern society.Wind energy is widely distributed in nature,so it can have a promising development perspective.Wind turbine blades are necessary constituent of wind turbines,whose function is to convert wind energy into electricity,and will be subject to variable extent of defects or damage during production,transportation operation.If these defects are not identified immediately,they may result in significant security accidents in the future.Consequently,an efficacious and implementable defect inspection technique is required to detect wind turbine blade defects.Based on the above,this thesis takes wind turbine blades as the object of study and uses electromagnetic ultrasonic technology as the detection method to study the internal defects of wind turbine blades.The main work content is divided into the following aspects.Firstly,according to the principle of electromagnetic ultrasonic,this thesis applied COMSOL finite element simulation software to establish the electromagnetic ultrasonic shear wave transducer model,and carried out the simulation analysis of electromagnetic ultrasound detection on the model of defect-free,bubble defect,inclusions defect and degumming defect of the blade.The simulation results show that the defect-free waveform and defect waveform can be distinguished effectively by comparing ultrasonic echo signals,which verifies the feasibility of using electromagnetic ultrasonic detection method for wind turbine blade defect detection.Secondly,With the modeling of the electromagnetic acoustic transducer,the COMSOL simulation software was used to study the factors affecting the transducer efficiency in terms of both permanent magnet structure and coil properties.The orthogonal test scheme was designed based on the 4 factors that have significant impact on the transducer.The orthogonal table with4 factors and 4 levels was made to optimize the structural parameters of the transducer,taking the peak value of Lorentz force as the index.The results indicate that the efficiency of the optimized Lorentz force peak is significantly higher than that of the initial parameter state.After the extreme difference analysis: the order of importance of the influence of each factor on the peak Lorentz force was obtained.Finally,in order to accurately and efficiently classify the ultrasonic echoes of defect-free,bubble,inclusion and degumming blades,variational mode decomposition combined with fuzzy entropy was used to extract the features of the defect echo data,and the extreme gradient boosting(XGBoost)multi-classification model was established.To improve the classification accuracy of the XGBoost model,particle swarm optimization(PSO)was used to optimize XGBoost hyperparameters.The defect sample data were brought into the PSO-XGBoost model to complete training and testing,and classification results were compared with decision tree,gradient boosting decision tree and random forest.The results show that the PSO-XGBoost multi-classification model established in this thesis has an accuracy of 96.7% for wind turbine blade defect identification with a small number of data samples,which proves the effectiveness and feasibility of the method used.