Research For Fault Diagnosis Of Wind Turbine Gearbox Based On Transient Time-series Symbolic Dynamics

In recent years,the development and utilization of the clean energy has become a hot research topic over the world for the pollution of the environment.In our country,wind power industry is developing rapidly by the support of the new energy policy and a lot of new results have been gained in the wind power industry of our country.However,compared with the other countries,the development space is rather broad such as the fault of the wind turbine,because of the industry starting lately and the immature technology.The investigation and analysis of many faults of the wind turbines show that the wind turbine gearbox is the one of the core parts which always leads to fault.The fault of the gearbox is often caused by overload in some severe environments and the repair work is hard to do.The loss and consumption are enormous.Therefore,it is very important to detect and diagnose the fault and to maintain the wind turbine operating reliably and stably.In this paper,the faults in the wind turbine gearbox has been analyzed at first,of which the reason and mechanism has been introduced and classified.The different faults,just as gear fault,bearing fault,shaft fault and so on,can be detected by the vibration signal,which is also analyzed in the following part by FFT in frequency domain to gain the spectrum characteristics.Then,since the non-stationarity and the spectrum leakage of the sampling vibration data in different rotation frequencies,a resampling method by the cubic spline interpolation and a time-domain average method are utilized to process the sampling data.All the data in different rotating frequency are uniformed to a constant frequency and eliminate some noise for fault diagnosing.The efficiency also been testified by some analysis in frequency-domain by FFT.This paper proposed a new fault diagnosis method by symbolic dynamics which named symbolic transient time-series analysis method.This method firstly transforms the sampling data to symbol sequence by partitioning,also called quantization.Thereafter,a probabilistic finite state automaton(PFSA)model is developed from the finite-length symbol sequence,which is a classifier for identification of the probability based on the transient data in both training and testing phases.After train the PFSA by the training data,the model can be used to identify any checking data by probability.At last the PFSA model based on symbolic dynamics has been testify in MATLAB that the relationship between the checking data length and the identification accuracy is also been analyzed.The calculating results show that this method by symbolic dynamics for fault diagnosis of wind turbine gearbox is effective and accurate.