【摘要】 针对最小熵解卷积(minimum entropy deconvolution, MED)在故障诊断中倾向于恢复少量伪主导冲击以及依赖经验选取滤波器长度的问题，提出了一种增强自适应盲解卷积方法。该方法设计一种非线性变换以增强滤波信号中的故障冲击特征，并将其融入滤波器系数的迭代求解中，从而解决MED因少量伪主导冲击造成峭度过大而无法有效恢复周期性故障冲击的问题。同时，所提方法提供一种可根据待分析信号自适应获得合适滤波参数的策略，进而克服传统依赖经验取值的不足。仿真信号与齿轮植入故障信号分析结果验证方法对于增强故障冲击及自适应选取滤波参数的有效性，实现周期性故障冲击的准确恢复。在列车齿轮故障诊断的工程实际案例中，所提方法准确诊断出齿轮传动系统中大齿轮的早期裂纹故障。与MED的对比研究，进一步表明所提方法在故障冲击增强与自适应恢复方面的优势。更多还原

【Abstract】 Here, aiming at the problem of minimum entropy deconvolution(MED) tending to recover a small amount of pseudo dominant shocks and rely on experience to select filter length in fault diagnosis, an enhanced adaptive blind deconvolution method was proposed. With this method, a nonlinear transformation was designed to enhance fault impact features in filtered signals, and it was integrated into iterative solving of filter coefficients to solve the problem of MED being unable to effectively recover periodic fault impacts due to excessive kurtosis caused by a small amount of pseudo dominant shocks. Meanwhile, the proposed method provided a strategy being able to adaptively obtain appropriate filtering parameters according to signals to be analyzed to then overcome shortages of traditional relying on experience. The analysis results of simulated signals and gear implanted fault signals verified the effectiveness of the proposed method to enhance fault impacts, adaptively select filter parameters and realize correct recovery of periodic fault impacts. It was shown that in practical engineering cases of train gear fault diagnosis, the proposed method can accurately diagnose early crack faults of big gear in gear transmission system; compared with MED, advantages of the proposed method in fault impact enhancement and adaptive recovery are further revealed.更多还原