The Diagnosis Of Early Stage Mechanical Fault For Wind Turbine Base On Single Channel Signal Blind Source Separation Method

As a result of the wind turbine working condition is usually abominable and astable.with the status changs of wind speed, power gird, temperature,the load case continuouslyadjusted too. so the load of the gear chain is continuosly changing. The reliability ofvarious parts should be guaranteed: The first, the parts quality, the second,if the parts growearly stage mechanical fault, it can be found. It is helpful for the equipment’s reliablyworking if we detect the fault in its early stage. For the first point,it related to the designand manufacture, the paper does not discuss, and We discuss the second point only.When the wind turbine get out of order, the fault components usually make a typicalfeature signal in running. Because the incipient fault signal is weak and usually submergedin background noise and it is difficult to be extracted. The background noise contain thestrong signal and other noise of the wind turbine auxiliary mechanic system. If the signalhas some noise, the blind source separation algorithm would not work it out properly.Therefore, it is important to get rid of the strong signal and denoise from the mixture signal.signal is denoised by autocorrelation method and EEMD algorithm. the strong signal isEliminated by extend virtual channel FastIca technology from the mixture signal.It isnecessary to satisfy the demand of BSS’s MIMO and extrac the incipient fault from thedata by EEMD-FastIca technology. The early stage knowledge needn’t be grasped toprocessing the data by these methods.It can process the single channel signal by BSS In thecondition of the unknown number of the source signal.The organic combinationapplication of these methods can improve the efficiency and the accuracy of diagnosis. Inorder to verify the validity of these methods, by simulating the typical vibration signals of the turbines mechanical system, whather or not the the weak signal can be effectivelyseparated from the mixed signal.By studying the frequency characteristic, An experiment is done by the wind turbinevibration diagnosis. by means of EEMD-FastIca algorithm, the characteristics signal issuccessful separated from the wind turbine test modal and the3MW wind turbine gearbox.the algorithm is effective in wind turbine signal processing;By means of virtual channelblind source separation techniques, the early fault signal is diagnosised and analysisedsuccessfully in a1.5MW wind turbine’s axial flow fan. thus the EEMD-FastIca algorithmand the virtual channel blind separation method and the extended algorithm is applicable towind turbines in signal processing,and it apply to the prediction of mechanical early systemfault.