Evolution characteristics of lag distance in fully mechanized mining face and hybrid deep learning prediction model

The cooperative operation state of the shearer and hydraulic supports in a fully mechanized mining face directly affects roof support safety and production continuity. Mismatch between shearer advancement and hydraulic support follow-up movement can expand the unsupported roof area and increase the risks of roof instability and support failure. To address problems such as abnormal increase of support-moving lag distance and difficulty in timely and accurate identification of lag distance categories when the shearer advancement rhythm does not match the hydraulic support follow-up movement rhythm during field operation, this paper took a fully mechanized mining face in a medium-thick coal seam in Inner Mongolia as the engineering background. The distribution characteristics of lag distance and its correlation with support pressure and shearer traction speed were analyzed, and lag distance was divided into three categories: small, normal, and large lag distance. A TCN-BiLSTM-Attention hybrid deep learning model was proposed. The TCN branch extracted local features from multidimensional time-series data and captured short-term fluctuation characteristics of support pressure. The BiLSTM branch mined long-term dependencies in time-series data and captured dynamic variation trends of lag distance under operating conditions. The Attention branch learned importance weights of feature dimensions and highlighted influence of key features. Through multi-branch feature fusion, the model realized zone prediction of different lag distance categories. Experimental results showed that overall prediction accuracy of the TCN-BiLSTM-Attention model reached 87.13%, and recall for high-risk large-lag-distance conditions reached 81.58%, outperforming similar models. Field application results showed that under model-guided control, the proportion of large lag distance occurrences and passive shutdown frequency decreased by 59.96% and 80.00%, respectively, providing effective support for adaptive cooperative control of shearer and hydraulic supports and reduction of roof support risk.