【摘要】 针对电动汽车齿轮箱齿轮故障特征背景噪声干扰大，难以提取的问题，提出了一种基于奇异谱分析和自适应最大二阶循环平稳盲解卷积(maximum second-order cyclostationarity blind deconvolution, CYCBD)的齿轮故障诊断方法。使用奇异谱分析(singular spectrum analysis, SSA)提取故障敏感成分；利用包络谐波乘积谱(envelope harmonic product spectrum, EHPS)对其进行估计以获得真实故障特征频率；基于自相关能量的效率评价指标(efficiency assessment index based on autocorrelation energy, EAAE)为依据，自适应选择CYCBD的滤波器长度；采用EHPS所得真实故障频率和EAAE自适应选择的滤波器长度进行CYCBD滤波，并对滤波后信号进行包络解调，从而进行齿轮故障诊断。通过对仿真信号、电动汽车传动系统故障测试试验台齿轮箱试验数据分析，结果表明：该方法可在无需先验知识情况下有效地去除背景噪声干扰、提取微弱齿轮故障脉冲，与其他自适应特征提取方法相比在故障频率提取效果和性能上更优。更多还原

【Abstract】 To address the challenges of extracting gear fault features with high background noise interference in electric vehicle gearboxes, a gear fault diagnosis method based on singular spectrum analysis(SSA) and adaptive maximum second-order cyclostationarity blind deconvolution(CYCBD) is proposed. Firstly, SSA is employed to extract fault-sensitive components. Next, the envelope harmonic product spectrum(EHPS) is used to estimate the true fault characteristic frequency. Then, an efficiency assessment index based on autocorrelation energy(EAAE) is introduced to adaptively select the filter length for CYCBD. Finally, CYCBD filtering is performed based on the estimated fault frequency from EHPS and the adaptively chosen filter length from EAAE, followed by envelope demodulation for gear fault diagnosis. By analyzing the simulated signals and the gearbox test data from the electric vehicle drivetrain fault test bench, the results show that the method can effectively remove the background noise interference and extract the weak gear fault pulses without a priori knowledge, and it is better in terms of fault frequency extraction effect and performance compared with other adaptive feature extraction methods.更多还原