Answers are grounded in this paper and cite their sources.
Predictive modeling study demonstrates accurate forecasting of hydraulic support resistance under thick roofs, indicating strong utility for real-time ground-pressure warning.
Thick sandstone roofs are widely distributed in the Ordos Basin. Under these conditions, strong ground-pressure events occur frequently and affect safe and efficient coal production. To address this problem, this study develops Ensemble-Frequency-Net (EFNet) for near-term prediction of hydraulic support working resistance. The model provides auxiliary information for ground-pressure warning under thick sandstone roofs. EFNet uses a bidirectional gated recurrent unit to extract temporal features. It also introduces Fast Fourier Transform, Short-Time Fourier Transform, and autocorrelation branches to extract frequency and lag features from the historical resistance sequence. The dataset contains 14,106 resistance records from six representative hydraulic supports with a sampling interval of 5 min. A 60 min historical resistance sequence is used to predict the working resistance over the next 25 min. EFNet achieves a mean absolute error of 0.864 MPa, a root mean square error of 1.847 MPa, and a coefficient of determination of 0.794. Under the investigated field datasets, the model shows statistically supported advantages over the comparison models in overall prediction accuracy and stability. Ablation results show that the FFT, STFT, and autocorrelation branches provide complementary information for resistance prediction. Additional evaluation on an adjacent working face, including direct zero-shot application without target-face weight updating, examines cross-face performance under comparable field conditions. The inference time of EFNet is 0.025 ms per sample, which is much shorter than the 5 min monitoring interval. The results show that combining temporal and time-frequency features improves the near-term prediction of hydraulic support working resistance. EFNet can therefore provide an auxiliary tool for ground-pressure monitoring and warning under the investigated thick sandstone-roof conditions.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: