Aiming at the problem of insufficient accuracy caused by the insufficient mining of spatiotemporal features in the process of unsafe behavior and danger identification of construction personnel, the traditional two-stream convolution model is improved, and a two-stream convolution dangerous behavior recognition model based on Faster R-CNN-LSTM is proposed. In this model, the Faster R-CNN network is connected in parallel with the LSTM network. The Faster R-CNN network is used as the spatial flow, and the human spatial motion posture is divided into static and dynamic features to extract the anchor point features, respectively. The fusion of the two is used as the output of the spatial flow. An improved sliding long-term and short-term memory network is used in the time flow to increase the extraction ability of the time series features of the construction personnel. Finally, the two branches are fused in time and space to classify and identify whether the construction personnel wear safety helmets. The results show that the MAP of the improved Faster R-CNN-LSTM network framework is increased by 15%. The original CNN-LSTM network framework detected four targets, but there was one misdetection, with an accuracy of 91.48%. The improved frame detection accuracy reaches 99.99%, and there is no error detection. The proposed method is superior to the pre-improvement and other methods that can effectively identify the unsafe behavior of construction workers on construction sites and also has a good distinction effect on fuzzy actions.

As coal mine production enters the deep mining stage, the impact of coal and rock dynamic hazards is becoming more and more significant. And the coal and rock containing initial damage such as fractures are more susceptible to destabilization damage by disturbance. So, this paper takes coal containing macro-crack with different inclination angles as the research object and uses the RMT-150B rock mechanics system to carry out uniaxial loading rupture tests on the specimens. On this basis, the changes in infrared radiation on the surface are observed using an infrared thermal imaging camera, and it is analyzed and studied according to the stress distribution and energy change of the specimens. The results show that the strain ratio at crack closure after bearing the coal gradually increases with the increase in the macro-crack inclination. When the inclination angle is 0° < α < 90°, there are obvious low-temperature bands on the upper and lower sides after macro-crack closure. The variance of the infrared thermal image of the specimen can reflect its infrared radiation information more effectively and has a good correspondence with the stress-strain curve. With the increase in the specimen macro-crack inclination angle, the linear change of VIRT is more obvious, the rate of change gradually increases, and the inclination angle is the maximum at 90°. The accumulated elastic strain energy U _e is the main source of energy for the sudden change in infrared radiation generated during the bursting process that occurs when the specimen is damaged, and U _e is linearly and positively correlated with the change in infrared radiation in front of the specimen peak. These will provide some experimental basis and theoretical guidance for the use of infrared radiation precursor characteristics to warn the damaged coal-rock dynamic disaster.

In order to study the precursor information of infrared radiation temperature of coal containing macro-cracks during load bearing damage, infrared monitoring tests of coal instability damage under different crack inclination angles are designed. Based on the critical slowing down theory, the variance and autocorrelation coefficients of the infrared radiation temperature series are calculated to reflect the damage precursors and provide a theoretical basis for early warning of coal and rock dynamic hazards. The results show that the infrared radiation sequences of pre-cracked coals have a critical slowing down effect. The autocorrelation coefficient and the variance of MIRT curve can be more accurate to find the precursor time of instability. In the critical slowing down analysis, the variance is more stable, and the lag step has a great influence on the autocorrelation coefficient. The infrared precursory information of the coal is mostly about 80% to 85% of the peak stress or instability time, and the horizontal fracture will make the failure precursory appear in advance.

The dust in the breathing zone of the fully mechanized mining face poses a serious threat to the lives and health of the workers, and clarifying the distribution and settlement pattern of the dust in the breathing zone is an important foundation for reducing dust concentrations in the breathing zone and improving the workers' working environment. The dust concentration distribution at the mining face and the settlement results of different dust particle sizes along the breathing zone were analyzed in detail using a combination of computational fluid dynamics simulation and downhole measurements by establishing a highly simulated 3D solid model of the 2302 fully mechanized mining face at the Xinqiao coal mine. The study's findings reveal that dust from advanced support and coal-cutting processes converge within 20-60 m of the coal mining region, generating a high-concentration pollution zone at the height of the sidewalk breathing zone. The particle size of dust in the breathing zone is primarily concentrated between 1 and 40 μm, and the fully mechanized mining face is split into a rapid settling area, a medium settling area, and a slow settling area based on particle size settling speed. Eddy currents in the return airway lead to an increase in dust particles of 20-40 μm in the breathing zone area. The results of the study can provide a theoretical basis for dust management at the fully mechanized mining face for the breathing zone area.

To further improve the accuracy of recurrent neural network in predicting the gas concentration in the upper corner of the mine tunnel, this paper proposes a method to construct a gas concentration prediction model based on multiple sequence long and short memory network, considering the spatial correlation between the gas concentration in the return airway and upper corner. The reliability of the model construction is improved by using the white noise test and smoothness test to verify the interpretability of the data in this paper and constructing supervised learning type data for gas concentration prediction model training and testing by means of data set division and data windowing. Through experimental comparison, grid search, and time series decomposition, the model algorithm, training parameters, and experimental results were combined to make an in-depth analysis of the influence of each parameter on the model training and the prediction. A training model of the spatially fused gas concentration prediction model with a network layer of 1 and a number of neurons of 32 as the model structure, Adam as the optimization algorithm, and a learning rate of 0.001 and a batch size of 32 as the training parameters was finally determined. The gas concentration prediction model trained in this paper performed well in the test set with a mean square error (MSE) of 0.0013, and its superiority was verified by comparing it with other models to provide some experience and basis for subsequent studies on gas concentration prediction in the upper corner.