Wind Turbine Gearbox Condition Monitoring System Based On Vibration Analysis

In face of the current situation of energy shortage and environmental pollution at home and abroad,wind energy as a clean and renewable energy has attracted people's attention.With the rapid development of wind power industry,the number of wind turbine assemblies has been increasing year by year.However,due to the harsh working environment,the transmission chain components of wind turbine are prone to be damaged.Therefore,once the wi nd turbine transmission chain is completed,a ground test system will be used to simulate the real situation.In this paper,a set of fault monitoring system based on vibration monitoring is developed for the wind turbine transmission chain of the ground test platform.Such a system is capable of pretreatment,vibration analysis,feature extraction and fault classification.The main respects of this paper are as follows:1.The requirements of the monitoring system were put forward and the preliminary design of the monitoring system was made.A piezoelectric accelerometer,a PCI-8814 data acquisition card and a next-turn industrial computer are chosen for hardware monitoring system Combined with the fault type and structure characteristics of the transmission chain of wind turbine,a layout for monitoring points was proposed.2.Low-pass filtering and wavelet threshold denoising methods are used as pre-processing(de-noising)methods for the system.Using different parameter combinations to denoise the experimental data,verifying the effectiveness of wavelet threshold denoising,and providing a method for finding the optimal solution of wavelet threshold denoising for the system.3.Time domain and frequency domain analysis are used as real-time analysis methods of the system,and statistical indicators are selected as their analysis parameters.Envelope demodulation analysis and wavelet packet decomposition are used as off-line analysis methods,and the characteristics of the two methods are verified by examples.4.According to the analysis method,three eigenvalue extraction schemes are proposed:time domain,frequency domain statistic eigenvalue extraction scheme,envelope demodulation eigenvalue extraction scheme and eigenvalue extraction scheme based on wavelet packet energy and reconstructed signal time domain index.The third scheme combines the advantages of wavelet packet energy to show the difference of spectrum distribution and the characterization of time domain statistic,and jointly extracts the energy of the wavelet packet with larger energy and the time domain characteristics of the reconstructed signal.The two sets of eigenvalues extracted from the experimental data were classified using a support vector machine(SVM).Under the same parameters,the classification result of the feature values extracted by the third scheme is better than the first scheme,which shows superiority.Based on the eigenvalue extraction scheme and the SVM classification method,the system fault diagnosis method is implemented.5.The monitoring system software was designed and developed to realize a series of functions,viz.hardware configuration,linear calibration of sensors,storage of vibration signals,preprocessing,analysis,eigenvalue extraction,classification and display.6.The acquisition accuracy and detection accuracy of the monitoring system are tested.By way of adjusting the sampling parameters,the sampling precision was found to be consistent with the sampling precision of the data acquisition card.The range of system detection errors is between-3.8%-2.19%,which meets the requirements for industrial application.