The control of hard roof conditions in underground coal mining is critical for ensuring mining safety and efficiency. Hard roof control remains a critical challenge in underground mining, particularly affecting mining safety and efficiency. Traditional support theories often show limitations in addressing the complex interactions between bolt spacing patterns and geological variability. This study focuses on the No. 10904 working face of Jingu Mine, where three distinct roof types are present: Type I (thick limestone roof, TLR, ≥1.2 m), Type II (moderate limestone roof, MLR, 0.5–1.2 m), and Type III (composite mudstone–limestone roof, CLR). Through FLAC3D numerical simulation and field validation, the mechanisms of bolt support under hard roof conditions were systematically investigated, and the optimization of bolt support parameters, including spacing, length, and pre-tension force, was conducted. The results indicate that: (1) when the ratio between lateral and row spacing approaches unity, reducing lateral spacing while increasing row spacing enhances support effectiveness, achieving 2 mm less roof subsidence with the 1.0 m × 1.5 m configuration compared to the 1.4 m × 0.8 m arrangement, despite a 21% reduction in bolt density; (2) an optimal rib bolt length of 1.8 m was determined, with support effectiveness diminishing beyond 2.0 m, and 1.5 m-long bolts reducing rib convergence by 15% compared to unsupported conditions; and (3) when the anchoring length exceeds 60% of the total bolt length, further increases in bolt length have minimal impact on deformation control under TLR, MLR, and CLR conditions. Field implementation of the optimized support scheme confirmed its effectiveness, with borehole television inspection showing no separation or fracturing within the monitored depth of 4 m in the roof strata. These findings provide practical guidelines for support design in similar geological settings, particularly for shallow-buried roadways with hard roof conditions.

Hydraulic supports in coal mining faces require continuous pressure monitoring to detect anomalies indicative of roof instability or equipment failure. Existing reconstruction-based methods rely on standard convolutional or recurrent encoders whose limited receptive fields or coarse temporal representations restrict detection accuracy; static per-window thresholding further discards temporal continuity during online deployment. This study proposes a temporal convolutional network autoencoder (TCN-AE) coupled with a Cumulative Sum (CUSUM) control chart for online anomaly detection in hydraulic support pressure data. The TCN encoder uses dilated convolutions with symmetric padding and residual connections, producing an exponentially expanding receptive field that captures temporal patterns at multiple scales. The CUSUM chart accumulates sustained positive deviations in the reconstruction error sequence, improving detection sensitivity while suppressing isolated false alarms. Component analysis experiments on synthetic anomalies show TCN-AE achieves an AUC of 0.811, outperforming CNN, LSTM, GRU, and fully connected autoencoder variants, along with Isolation Forest and One-Class SVM. On a manually curated real fault test set, where per-window reconstruction scores carry negligible discriminative information (AUC = 0.586, near chance), the CUSUM strategy exploits temporal continuity to improve F1 from 0.213 to 0.905 for TCN-AE. This +0.692 gain is driven entirely by temporal accumulation rather than model discriminability, demonstrating that the CUSUM framework is most valuable precisely when per-window signals are weakest.