【摘要】 针对数据驱动时频分析方法(Data-Driven Time-Frequency Analysis,DDTFA)的初始相位函数估计直接影响算法的收敛性及分解精度的问题,将多尺度线调频基稀疏分解方法(Multi-Scale Chirplet Sparse Decomposition,MSCSD)引入DDTFA的初始相位函数估计中,提出了MSCSD-DDTFA方法,并应用于变转速齿轮故障诊断中。MSCSD方法采用分段线性拟合的思想,可从低信噪比信号中精确地估计出信号的瞬时频率,进而求取相位函数;DDTFA方法则可根据MSCSD估计的相位函数不失真地分离出时变非平稳信号分量;最后,可根据MSCSD估计出的瞬时频率对信号分量进行阶次包络分析,获取阶次包络谱以诊断变转速齿轮故障。算法仿真和应用实例表明:该方法可准确分离出信号中的时变非平稳信号分量,并提取变转速齿轮故障特征。更多还原

【Abstract】 Aiming at the initial phase function estimation for data-driven time-frequency analysis(DDTFA)method,which directly influences the convergence and decomposition accuracy of the algorithm,a multi-scale chirplet sparse decomposition(MSCSD)method which utilizes the idea of piecewise linear fitting is introduced to estimate the initial phase function of DDTFA,and thus the MSCSD-DDTFA method is proposed accordingly and applied to a gear fault diagnosis under variable rotating speed.The instantaneous frequency can be accurately estimated from signal with low signal-noise ratio by using MSCSD,based on which the phase function can then be obtained,and then the time-varying non-stationary signal component can be separated from the signal without distortion by DDTFA according to the estimated phase function via MSCSD.Lastly,according to the estimated instantaneous frequency via MSCSD,the order envelope spectrum is carried on to the signal component so as to diagnose the gear fault under variable rotational speed.Simulation and experiment show that the proposed method can accurately separate time-varying non-stationary signal component and extract the gear fault characteristic from the signal.更多还原