Research And Implementation Of Fault Diagnosis Of Wind Turbine Blades

Wind turbine blade is one of the key components of wind turbine, and with the increase of unit capacity, the length of blade becomes longer and the distance to the ground is larger. Because of those, the probabilities of damages become bigger. More and more scholars pay attention to the study of the condition monitoring and damage detection of wind turbine blade. Because the blade is located in the outside of cabin and be in a suspended state. Wind turbine was forced to stop when the damage was happening, which can led to the power generation efficiency lower. It will cause serious loss of manpower and material resources when failure has happened, and it will need more maintenance costs. So the research on the condition monitoring and damage identification of wind turbine blade becomes more and more important. The thesis puts forward a thought of detection system that the acoustic emission signal collection was based on acoustic emission technology and damage identification was based on support vector machine(SVM). The crack and edge damage acoustic emission signal was collected by hardware experiment, and the signal was transmitted to PC. The signal will be processed after the PC receives the signal, and the SVM prediction model will be established and optimized. Then the prediction model with higher classification accuracy rata is validated in the monitoring system. The main research content is as follows:(1) The damage acoustic emission signal is collected by using hardware experiment. The thesis selects the acoustic emission technology to collect signal. Then the signal is send by wireless transmission module after preamplifier amplification and modulus conversion. Those parts are put on the blade. The wireless module which is located in cabin of wind turbine receives the signal and sends it to the PC on the ground by RS485. The thesis do static loading experiment with small glass fiber reinforced plastic materials after the experiment equipments assembled and debugged completely. The thesis makes the blade produce acoustic emission signal and collects it by artificial simulation the crack and edge damage of wind turbine.(2) The PC completes the signal receiving, processing and establishes the support vector machine prediction model. The PC interface is programmed by VB language. And it uses own serial communication function to receive signal and uses the ActiveX automation protocol technology to call MATLAB to process the signal. First of all, the thesis normalizes the signal data, then decomposes the dates and extracts the feature information by using db wavelet, and then the SVM model is established by using the feature information.(3) The parameters of SVM is optimized by particle swarm optimization algorithm(PSO) and fruit fly optimization algorithm(FOA) respectively. Because of the fast convergence speed and strong optimization ability of both the PSO and FOA, so the thesis uses two kinds of optimization algorithm to optimize the SVM model and selects the model with higher classification accuracy as the damage identification model.(4) The thesis verifies the damage identification model in detection system. It completes the damage identification in the detection system after the optimized model with higher accuracy is selected. And the thesis verifies the accuracy of damage identification and feasibility of the detection system.The results show that the optimized damage identification model of support vector machine is available and the detection system of damage identification of wind turbine blades is feasible. The research in the thesis which is done under the condition of laboratory can provide a method for following study.