基于变分模态分解和粒子群算法的微震信号降噪方法

为从含噪微震信号中提取有效信息, 并准确识别岩体破裂信号和爆破振动信号, 提出了基于粒子群算法和小波阈值去噪的改进变分模态分解方法。该方法利用粒子群算法实现模态数量和惩罚因子的最优取值, 以最优参数对微震信号进行变分模态分解, 再对由高频噪声主导的模态分量进行小波阈值去噪, 将去噪后的高频信号分量与原先的低频信号分量进行重构, 实现信号降噪。经验证, 该方法相比集合经验模态分解和单纯的变分模态分解方法具有更好的降噪效果。以该方法对200组岩体破裂信号和200组爆破振动信号进行去噪, 以第一模态分量能量占比50%作为区分爆破振动信号和岩体破裂信号的依据, 识别成功率达到97.25%, 证实了此识别方法的准确性。

Noise Reduction Method for Microseismic Signal Based on Variational Mode Decomposition and Particle Swarm Algorithm

In order to extract effective information from noisy microseismic signals and accurately identify fracture signals of rock mass and blasting-induced vibration signals, an improved variational mode decomposition (VMD) method based on particle swarm optimization (PSO) and wavelet threshold denoising was proposed. This method can employ PSO to obtain the optimal value for the number of modes and the penalty factor. The microseismic signal is subjected to VMD with optimal parameters. Then the wavelet threshold denoising is performed for the modal components dominated by high-frequency noise, and the high-frequency signal components after noise reduction are reconstructed with the original low-frequency signal components, so as to realize noise reduction of signals. It has been verified that this method has better noise reduction effects than ensemble empirical mode decomposition (EEMD) and simple VMD. Later, 200 sets of signals about rock mass fracture and 200 groups of signals about blasting vibration were denoised using this method. The signals about blasting vibration was distinguished from the signals about rock mass fracture based on the energy of the first modal component accounting for 50%, showing the recognition rate up to 97.25%. It is confirmed that the recognition with this method is accurate.