Micro-seismic resolution in mining engineering by combining denoising processing and improved CNN algorithm

Microseisms in mining engineering may not only damage the structure of mines, but also interfere with their production activities and induce other geological disasters. Therefore, accurately distinguishing micro-seismicity in mining engineering is of great significance for ensuring mine safety and preventing geological disasters. To distinguish microseisms in mining engineering, a signal denoising method grounded on variational mode decomposition (VMD) algorithm and permutation entropy was studied and designed, and sparrow search algorithm was introduced to optimise the parameters of VMD algorithm. High-quality input data foundation for subsequent micro-seismic resolution models was provided through this denoising method. Subsequently, a micro-seismic resolution model combining transformer and convolutional neural network was developed, which utilises transformer to focus on important information and obtains feature information through depthwise separable convolution. The findings denoted that the designed denoising method achieved maximum signal-to-noise ratios of 33.142 dB, 34.021 dB and 33.743 dB on simulated signals from Blocks, Doppler and Heavyisine, respectively, all of which were higher than the comparison method. The average root mean square error of this method in practical applications was 3.088 × 10 −6 . The accuracy and maximum root mean square error of the micro-seismic resolution model were 97.54% and 2.95 × 10 −6 , respectively. The average time consumption and F1 score were 7.12 ms and 0.9456, which were better than the comparison model. On the training set, the model correctly identified 491, 497, 493, 508 and 498 micro-seismic information, rubber hammer vibration information, iron hammer vibration information, excavation vibration information and blasting vibration information, respectively, which were closer to the true values. The designed noise reduction method and resolution model have good effects and can provide accurate signal processing and analysis tools for micro-seismic resolution in mining engineering. The novelty of the research lies in the combination of micro-seismic signal denoising and micro-seismic identification resolution, which avoids the uncertainty caused by manually setting parameters and comprehensively improves the resolution accuracy of micro-seismic research in mining engineering, surpassing previous research efforts in micro-seismic monitoring accuracy in mining engineering.