基于粒子群优化匹配追踪的风机振动信号去噪处理

摘要

Aiming at the problem that in the process of wind turbine vibration signal collection it is easily to be affected by noise,we proposed an over-complete atom dictionary-based matching pursuit algorithm to process the wind turbine vibration signal.The algorithm can adaptively extract the signal structure correlated with the atom,so that the noise suppression can be achieved.And in the process of matching pursuit algorithm operation,it uses an improved particle swarm optimisation algorithm in combination with the gradient information to search optimal atoms.Simulation result showed that the proposed algorithm had higher computation efficiency and signal reconstruction accuracy than the basic matching pursuit algorithm.Moreover,based on this algorithm,we carried out the experiment of denoising processing on the vibration signal of wind turbine gearbox,experimental result showed that the signal-to-noise ratio of the denoised signal could be improved to 5 dB and higher,the waveform features were more clearly,and while denoising,the fault information could be effectively reserved as well.%针对风力机振动信号采集过程中易受噪声影响的问题，提出基于过完备原子库的匹配追踪算法对风机振动信号进行处理。该算法能自适应提取和原子相关的信号结构，从而可实现噪声抑制。在匹配追踪算法处理过程中，利用结合梯度信息的改进的粒子群优化算法来寻找最佳原子。仿真结果表明，该算法比标准匹配追踪算法具有更快的运算效率及更高的重构精度。利用该算法对风力发电机齿轮箱振动信号进行去噪处理实验。实验结果表明，去噪后信号信噪比可提高5 dB以上，波形特征更加清晰，并且可以在降噪的同时有效保留故障信息。