Abstract: Aiming at the problem that the traditional multi-wavelet neighbor coefficient denoising method used the uniform threshold method to obtain the threshold accuracy was not high and the signal denoising was not ideal,a multi-wavelet neighbor coefficient denoising method based on hybrid particle swarm optimization was proposed.In this method,tabu search algorithm with global optimization ability and particle swarm optimization algorithm were fused,and the fusion algorithm was introduced into the multi-wavelet neighbor coefficient denoising method to improve its way of calculating the threshold.By comparing the noise reduction results of the traditional multi-wavelet adjacent coefficient denoising method,the noise reduction algorithm based on empirical mode decomposition and the multi-wavelet denoising method based on particle swarm optimization and genetic algorithm,the results showed that the multi-wavelet adjacent coefficient denoising method based on particle swarm optimization and tabu search algorithm could not only extract more fault feature information,but also enhance the fault frequency impact characteristics and effectively suppress the background noise.

Key words: multi-wavelet neighbor coefficient, hybrid particle swarm optimization algorithm, fault signal, denoising

摘要： 针对传统多小波相邻系数去噪法沿用统一阈值方法获取的阈值精度不高而导致信号去噪不理想的问题,提出了基于混合粒子群优化的多小波相邻系数去噪方法。该方法将具有全局寻优能力的禁忌搜索算法和粒子群优化算法相融合,并将这种融合算法引入到多小波相邻系数去噪方法之中对其阈值求取方式进行改进。通过对比传统的多小波相邻系数去噪方法、基于经验模态分解的降噪算法及基于粒子群和遗传算法的混合粒子群算法优化的多小波去噪方法降噪结果,结果表明：基于粒子群和禁忌搜索的混合粒子群算法优化的多小波相邻系数去噪方法不仅能提取更多的故障特征信息,还能够增强故障频率冲击特性和有效抑制背景噪声。

关键词: 多小波相邻系数, 混合粒子群优化算法, 故障信号, 去噪