【作者】 王雪；

【导师】 胡爱军；

【作者基本信息】 华北电力大学 ， 机械电子工程， 2014， 硕士

【摘要】 作为风力发电机的重要传动部件，齿轮箱是风力发电机组中故障率较高的部件之一，因此研究风力发电机齿轮箱故障诊断方法尤为重要。本文首先介绍了风力发电的国内外发展状况及风力机齿轮箱故障诊断方法的研究现状，阐述了时域指标诊断方法的基本理论，简述了去趋势波动分析及多重分形去趋势波动分析的基本原理。针对去趋势波动分析方法中区间长度选取的随机性和主观性，根据互信息的基本理论，提出采用符号分析的方法进行互信息计算，对去趋势波动分析的长程相关性指数进行优化，避免了区间长度对长程相关性指数的影响，确保了去趋势波动分析改进算法的稳定性。然后根据信号故障时会出现多重分形的特征，通过计算分形图偏离程度，得到图形偏离程度（DGD）作为故障识别的时域指标。进行了滚动轴承的信号仿真，对仿真信号进行去趋势波动分析改进算法运算，得到仿真信号的分形图和DGD特征因子，根据所得分形图和DGD特征因子来分辨正常与故障仿真信号。通过对仿真信号的分辨来证明了DGD特征因子的有效性和准确性。在仿真信号中加入不同程度的噪声，验证噪声对DGD特征因子的影响，证明了噪声的加入不影响DGD特征因子对故障识别。最后将去趋势波动分析改进算法应用到实际的信号中，对齿轮箱振动信号和风力机齿轮箱振动信号进行分析，分别求取其分形图和DGD特征因子，并将DGD特征因子与其他时域指标进行对比。通过对比发现，峭度在对实际故障识别的过程中出现了不稳定性，而DGD特征因子对风力机齿轮箱故障识别上有很好的稳定性，证明了本文提出的DGD特征因子对风力机齿轮箱故障识别上的优越性。更多还原

【Abstract】 As an important transmission components of wind turbine, gear box is one of thehigh failure rate of wind turbine components, so the fault diagnosis method of windturbine gearbox is particularly important.This paper first introduces the situation of wind power and the research status ofwind turbine gearbox fault diagnosis methods, this paper expounds the basic theory oftime domain index diagnosis method, introduced to Detrended fluctuation analysis andMultifractal Detrended Fluctuation. for the randomicity and subjectivity of selectinginterval length of Detrended fluctuation analysis. According to the basic theory of mutualinformation, adopting the method of analysis of symbols in calculating mutualinformation and optimizing the long-range correlation index of Detrended fluctuationanalysis in order to avoid the influence of interval length on the long-range correlationindex and to ensure the stability of the algorithm.Then according to the signal failure occurs multifractal characteristics, bycalculating the fractal figure deviation degree, get The Degree of Graphic Deviation(DGD) as an index of fault recognition time domain. Going the rolling bearing signalsimulation, operating the simulation signal by Detrended fluctuation analysis algorithm,get the simulation signal characteristics of fractal graph and DGD characteristic factor,according to the characteristics of the scoring charts and DGD characteristic factor todistinguish between normal and fault simulation signal. Through the simulation signalresolution to prove the effectiveness of DGD characteristic factor and the accuracy.Adding different levels of noise in the simulation signal, verifing the influence of thenoise characteristics of DGD characteristic factor,which proves that the addition of noisewill not affect the DGD characteristics factor of fault recognition.Finally the improved algorithm of Detrended fluctuation analysis is applied to theactual signal fluctuation analysis, analyzing the vibration signals of gearbox and windturbine gearbox vibration signal, to calculate the fractal graph and DGD characteristicfactor respectively, and characterizing DGD characteristic factor compared with othertime domain index. Found by comparison and kurtosis in appeared in the process ofactual fault identification of instability, and DGD characteristic factor on the factor ofwind turbine gearbox fault recognition has a good stability, this paper demonstrates thatthe proposed DGD characteristic factor of the superiority of the wind turbine gearboxfault recognition.更多还原