论文部分内容阅读
针对局域波分解后去除含噪分量与虚假分量的随意性和盲目性,提出了局域波互信息降噪方法。首先用局域波对信号进行分解,计算各IMF分量与原信号的互信息,再选取互信息值高的分量进行自相关分析去除噪声成分,并用有效分量对信号进行重构。为了验证该方法的有效性,运用仿真信号进行分析,并将其应用于齿轮箱故障诊断中。对局域波互信息降噪后的齿轮箱振动信号进行小波变换,并提取小波奇异谱熵作为故障特征量进行故障识别。将结果与未进行降噪处理的识别结果相对比,证明了局域波互信息降噪在工程实践中的实用性。
Aiming at the randomness and blindness of eliminating noisy components and false components after the local wave decomposition, a local noise reduction method based on mutual information is proposed. Firstly, the signal is decomposed by local wave to calculate the mutual information between each IMF component and the original signal. Then the component with high mutual information value is selected to perform the auto-correlation analysis to remove the noise component and reconstruct the signal with the effective component. In order to verify the effectiveness of the method, the simulation signal is used to analyze and apply it to gear box fault diagnosis. Wavelet transform is performed on the vibration signals of the gearboxes after the local signal noise reduction, and the wavelet spectral singularity entropy is extracted as the fault feature to identify the fault. Comparing the result with the recognition result without noise reduction, it proves the practicability of local wave mutual information noise reduction in engineering practice.