Methane is one of the most dangerous gases encountered in the mining industry. During mining operations, methane can be broadly classified into three states: mining excavation, stoppage safety check, and abnormal methane concentration, which is usually a precursor to a gas accident, such as a coal and gas outburst. Consequently, it is vital to accurately predict methane concentrations. Herein, we apply three deep learning methods—a recurrent neural network (RNN), long short-term memory (LSTM), and a gated recurrent unit (GRU)—to the problem of methane concentration prediction and evaluate their efficacy. In addition, we propose a novel prediction method that combines classical time series analysis with these deep learning models. The results revealed that GRU has the least root mean square error (RMSE) loss of the three models. The RMSE loss can be further reduced by approximately 35% by using the proposed combined approach, and the models are also less likely to result in overfitting. Therefore, combining deep learning methods with classical time series analysis can provide accurate methane concentration prediction and improve mining safety.

In this study, to further improve the prediction accuracy of coal mine gas concentration and thereby preventing gas accidents and improving coal mine safety management, the standard whale optimisation algorithm’s (WOA) susceptibility to falling into local optima, slow convergence speed, and low prediction accuracy of the single-factor long short-term memory (LSTM) neural network residual correction model are addressed. A new IWOA-LSTM-CEEMDAN model is constructed based on the improved whale optimisation algorithm (IWOA) to improve the IWOA-LSTM one-factor residual correction model through the use of the complete ensemble empirical model decomposition with adaptive noise (CEEMDAN) method. The population diversity of the WOA is enhanced through multiple strategies and its ability to exit local optima and perform global search is improved. In addition, the optimal weight combination model for subsequence is determined by analysing the prediction error of the intrinsic mode function (IMF) of the residual sequence. The experimental results show that the prediction accuracy of the IWOA-LSTM-CEEMDAN model is higher than that of the BP neural network and the GRU, LSTM, WOA-LSTM, and IWOA-LSTM residual correction models by 47.48%, 36.48%, 30.71%, 27.38%, and 12.96%, respectively. The IWOA-LSTM-CEEMDAN model also achieves the highest prediction accuracy in multi-step prediction.

Gas accidents threaten the safety of underground coal mining, which are always accompanied by abnormal gas concentration trend. The purpose of this paper is to improve the prediction accuracy of gas concentration so as to prevent gas accidents and improve the level of coal mine safety management. Combining the LSTM model with the LightGBM model, the LSTM-LightGBM model is proposed with variable weight combination method based on residual assignment, which considers not only the time subsequence feature of data, but also the nonlinear characteristics of data. During the data preprocessing, the optimal parameters of gas concentration prediction are determined through the analysis of the Pearson correlation coefficients of different sensor data. The experimental results demonstrate that the mean absolute errors of LSTM-LighGBM, LSTM and LightGBM are 1.94%, 2.19% and 2.77%, respectively. The accuracy of LSTM-LightGBM variable weight combination model is better than that of the two above models, respectively. In this way, this study provides a novel idea and method for gas accident prevention based on gas concentration prediction.

Intense cutting-induced dust production in fully mechanized mining faces (FMMFs) with large mining heights produces a high amount of dust that is difficult to capture and severely affects the working environment, threatening the health of occupational staff. The effective spray range and atomization performance of tracking sprays are counteracted by the influences of the mine’s height and ventilation airflow in FMMFs. Thus, optimizing the spray’s parameters and relationship between the effective spray range and atomization performance to reduce dust levels is the main priority of dust-control techniques. In this study, a new swirl-core atomization nozzle is developed based on fluid mechanics and the solid–liquid coalescence mechanism. The liquid generates a circumferential velocity when passing through the swirl core, which considerably increases the droplet breaking power and reduces the droplet cohesion factor, achieving a remarkable atomization effect. The spray angle of the new nozzle is 57°, which is 80.9% greater than the GZPW-16 mine-use nozzle (31.5°); the effective spray range increases from 5.2 to 5.9 m; and the spray’s mist saturation is significantly better than the GZPW-16 mine-use nozzle. Under different test pressures, the particle size range of the droplets produced by the new nozzle and dust particles on site satisfied the best synergy of droplet–dust coalescence. The total and respirable dust-reduction rates were 78% and 75.1%, respectively, which were 42% and 65% higher than those of the original nozzle. The new nozzle effectively improves the efficiency of the single dust-control technique of the tracking spray, which is significant for the dust-prevention and -control technology of FMMFs with large mining heights.

Underground gas sensors as the most intuitive tool for monitoring gas concentrations in underground mining, yet they are subject to frequent anomalies due to ground pressure, constructions, even malicious masking by workers. Due to the depth of underground mining and the complexity of the environment, it is almost impossible to manually monitor the status of the all the sensors. Thus, the ability to accurately identify the working status of gas sensors at the working face are critical importance to mining safety. In this paper, we propose a deep learning feature engineering based approach to coupling the relationship between underground sensors. Experiment results show that the relationship between gas sensors can be expressed by position and time, so that when a sensor such as upper corner T0 malfunctions, it can be detected by other sensors such as T1 and T2. By converting the gas concentration into the form of recurrence plots (RPs), we are able to transform time-series gas concentration data into images with more dimensions in time lag, and enabling the application of more efficient and accurate machine vision methods. Based on the location of sensors at the working face, we found that the sensors at positions T0, T1 and T2 are correlated as the wind flows through the tunnel and have a higher correlation in the subsections of the time series. And those correlation can directly use to check the operating status of the sensors. We also discuss whether the relationships between the data itself can be preserved at the feature level during the mapping of gas concentrations to features, since deep learning (DL) looks like the next promise future after digitization in the mining industrialization with more and more data analysis and placing the results under a larger decision. This feature-based approach for gas concentration analysis can also be used for prediction and early warning.