Since geological structures around coal mines are complex, sudden coal mine water inflow is seriously threatening coal mining safety. To improve the accuracy of predicting coal mine water inflow, a multi-source dataset is collected to develop a coal mine water inflow prediction model based on multi-factor Pearson correlation analysis, where a convolutional neural network and bidirectional long short-term memory neural network are adopted to extract features from time-series data. To validate the performance of the present prediction model, a case study is conducted, where the predicted coal mine water inflow is close to the collected coal mine water inflow. Meanwhile, compared to other prediction models, the present prediction model can predict the magnitude and development trend of coal mine water inflow in the next 8 h more accurately, where the mean absolute percentage error is 5.76% and the correlation coefficient is 0.922.

Planned cutting is a core technique for intelligent coal mining, relying on high-precision geological models of fully mechanized mining faces to plan the cutting trajectory of mining equipment, with model accuracy as a prerequisite for intelligent mining. To address the limitations of traditional interpolation methods in dynamic model updating and the technical gap between geological information and equipment control parameters, this study proposes a coal mining machine cutting path control method based on dynamic geological models. An improved smooth discrete interpolation method is developed to realize dynamic updating of the geological model, effectively improving the accuracy of local geological models and ensuring safe mining operations. Meanwhile, a method for converting geological information into coal mining equipment control parameters is proposed, breaking the technical barrier between geological data and production control information and laying a foundation for unmanned and intelligent mining. Field tests conducted in a shaft coal mine in Shaanxi demonstrate that the method achieves precise control of the coal mining machine’s trajectory: during a 7-day trial, the working face advanced 56 m and mined 51,000 tons of coal with minimal human intervention. Comparative analysis shows that the error between the planned cutting based on the dynamic geological model and manual cutting is within 10 cm, and the drum height curve is smoother, reducing frequent adjustments and facilitating equipment protection. Dynamic model updating ensures high accuracy, with an average absolute error of 0.029 m at 5 m from the update point and 0.101 m at 10 m, meeting the requirements for automated cutting. The successful application of this method verifies its feasibility in actual mining processes, providing a new technical approach for achieving unmanned and intelligent coal mining.

To reduce the risk of coal mine water inrush, a dynamic escape path optimization model based on the spatio-temporal evolution of the water inrush is studied. The actual coal mine is simplified into roadway nodes and segments to meet the real-time simulation of the coal mine water inrush, where the computational cost is reduced significantly while the accuracy is acceptable. To solve the control equations of the open channel flow and full channel flow efficiently, the lattice Boltzmann method is adopted to simulate the spatio-temporal evolution of the water inrush. Different from the previous studies, the spatio-temporal evolution of the water inrush is taken into account, which is closer to the actual case. The escape speed is not static, which is affected by the water depth dynamically; meanwhile, the effect of the physical energy reduction is considered. To validate the dynamic escape path optimization model based on the spatio-temporal evolution of the coal mine water inrush, three case studies are conducted. In the first case, there is one water inrush point and one person, while in the second case, there are two water inrush points and four persons; the third case is an actual coal mine with multiple water inrush points. We defined two indicators to evaluate the risk of the escape path quantitatively; they are the window escape time and rescue priority. By conducting the dynamic programming of the escape path, the optimal escape path is selected, where the effectiveness of the dynamic escape path optimization model is validated. The present work is helpful in reducing the risk of coal mine water inrush and improving the safety of the early warning system.