Intelligent phase picking of microseismic signals based on ResUNet in underground engineering

• ResUNet integrates ResNet's residual learning with UNet's multi-scale feature extraction, effectively mitigating vanishing gradients and enabling cross-layer fusion of shallow and deep features. • The model outperforms traditional STA/LTA algorithms and advanced networks like PhaseNet and EQTransformer in P- and S-wave phase picking accuracy, especially under low SNR conditions. • Trained on 400,000 labeled microseismic signals from the STEAD dataset, ResUNet demonstrates high robustness and generalization when applied to real-world microseismic monitoring at the Shizhuyuan polymetallic mine. • The proposed method provides reliable technical support for early warning of rockburst and other underground hazards, enhancing safety in deep underground engineering.