基于奇异值差分谱分析和蚁群算法的小波阈值降噪

摘要

针对传统小波阈值去噪阈值选取的问题,将奇异值差分谱的方法与蚁群算法相结合运用到小波阈值降噪中,提出一种小波系数双阈值寻优方法.首先将待处理含噪信号进行多尺度小波分解;之后根据每级信号小波系数的奇异值差分谱分析得到寻优的目标函数;然后根据目标函数利用蚁群算法在每级的小波系数上进行阈值寻优;最后重构经过最优阈值量化规则处理的小波系数得到降噪信号.通过对仿真信号的降噪处理表明本方法对不同特点信号的降噪效果要好于传统阈值降噪方法;对滚动轴承以及深沟球轴承的振动故障信号的降噪处理验证了方法的可行性和适用性.%In order to solve the problem of traditional wavelet threshold de-noising threshold selection,the singular value difference spectrum and the ant colony optimization(ACO) are applied to the wavelet threshold denoising,a method of double threshold searching is proposed.Firstly,the multi-scale wavelet decomposition was performed on the noisy signal to be processed.After that the optimization of the objective function was obtained based on the singular value difference spectrum analysis of the wavelet coefficients of each level signal.Then the threshold was optimized on each level of wavelet coefficients using the ACO based on the objective function.Finally,the denoising signal was obtained by reconstructing the threshold quantified wavelet coefficients.The noise reduction of the simulation signal shows that this method is better than the traditional threshold denoising method.The feasibility and applicability of the proposed method are verified by the noise reduction of vibration fault signals of rolling bearing and deep groove ball bearing.