In order to explore the problems of rescue evacuation and mechanical ventilation in the long-distance, large longitudinal slope and joint conversion of railway tunnels, taking the deep-burned and long-distance combination of railway tunnels-Zhujiangkou Tunnel as an example, the evacuation simulation software Pathfinder was used to simulate the evacuation time in the accident situation of the shield section, mining section and open-dig section of the Zhujiangkou Railway Tunnel, and at the same time, the corresponding rescue evacuation path and mechanical ventilation measures were formulated based on the calculation results. The research results indicated that among the three different construction methods, the shield tunnel section had the weakest capacity for evacuation, and the difficulty of personnel evacuation was relatively high. Evacuation could be carried out simultaneously through the track level rescue channel and the underground evacuation corridor; The inclined shaft in the mining section had the strongest evacuation ability as an emergency exit. Therefore, this project combined the characteristics of each section of the tunnel construction method and adopted a zoning rescue and evacuation method to shorten the tunnel and evacuate from multiple points. According to the calculation results of evacuation time, the necessary safe evacuation time for non fixed point evacuation inside the tunnel was greater than the available safe evacuation time, which could not meet the requirements for evacuation of personnel on fire trains. In case of tunnel fire, the train should be towed outside the tunnel for evacuation and rescue, and only the evacuation and rescue requirements for personnel under train fault conditions were considered inside the tunnel. Emergency exits and shelters were designed for ventilation according to train fault conditions. The ventilation of the evacuation corridor in the shield tunnel section adopted a mechanical air supply method at both ends, and the air supply volume was calculated based on the door opening wind speed method to meet the needs of fresh air for personnel evacuation. Based on the characteristics of crowd evacuation, measures were taken to open 4+1 protective doors near the end of the accident train to overcome the problem of air supply in long-distance evacuation channels. The research results can provide certain reference and ideas for the ventilation design of construction method combined railway tunnels under evacuation and accident conditions.

To scientifically determine the timing of hydraulic tunnel support under complex geological conditions in the southwestern region, the diversion tunnel of Xulong Hydropower Station was selected as the research object. Three-dimensional laser scanning, acoustic wave testing, and microseismic monitoring technologies were integrated into a collaborative analysis method for surface deformation, shallow damage, and internal fractures. Through continuous spatiotemporal monitoring, the time-dependent deformation characteristics of surrounding rock, evolution patterns of wave velocity fields, and spatiotemporal distribution modes of microfractures were systematically tracked. Monitoring data revealed that significant deformation of surrounding rock, wave velocity variations, and microseismic activities were observed during the initial stage. These parameters gradually stabilized 11 days after excavation. During this period, the self-bearing capacity of the surrounding rock was maximized, serving as the basis for optimal support timing. Engineering practice verified that the self-supporting capacity of Class II surrounding rock was fully utilized when supports were implemented 11 days post-excavation. This finding provides important references for determining support timing in similar projects.

This research was based on the actual working conditions of buildings in the Jiangnan section crossed by the Shangyuanmen Cross River Tunnel. A three-dimensional numerical simulation finite element model was established to simulate the gradual formation loss process after tunnel construction through displacement control method. After the simulation was completed, the settlement and horizontal displacement of the soil and frame structure, as well as the displacement and strain of each compartment of the frame structure, were extracted to analyze the displacement and strain distribution and development law of the frame structure caused by large-diameter tunnel construction. The research results showed that the settlement of the foundation of building frame structural plates decreased with increasing distance from the tunnel axis, while the horizontal displacement remained approximately at a stable value, which was smaller than the horizontal displacement of the underlying strata; The horizontal displacement distribution law of the frame structure was related to the tunnel crossing relationship. The horizontal displacement of the upper part of the building directly penetrated by the tunnel was smaller than that of the lower part, and the horizontal displacement of the upper part of the building laterally penetrated by the tunnel was larger than that of the lower part. However, the settlement of the frame structure changed relatively little in the vertical direction; The total vertical displacements of frame structure compartments increased with the increase of tunnel volume loss rate, and was mainly composed of shear displacement; The shear strain also increased with the increase of the volume loss rate of the tunnel, and the shear strain of buildings that were penetrated by the tunnel or buildings that were close to the tunnel were generally larger. At the same time, for multi compartment penetrating buildings, the shear strain along the compartment variation curve approximated a sine cosine curve.

In order to improve the safety of subway tunnel operation and provide a design basis for the planning of adjacent foundation pits, the deformation characteristics of existing subway segments and the surrounding soil pressure distribution under excavation unloading at different spatial locations were studied using a similar model test method based on Line 9 in the collapsible loess stratum in Xi'an. The influence zones of tunnel convergence displacement under the horizontal net distance between the foundation pit and the tunnel, the width, and the depth of the foundation pit were analyzed, and a modified formula of surrounding soil pressure was proposed. The results showed that the surrounding soil pressure of the shield tunnel exhibited a gourd-shaped symmetrical distribution before the excavation of the side foundation pit, with higher earth pressure at the top and bottom of the shield tunnel and lower pressure at the waist. After the excavation of the side foundation pit, the surrounding soil pressure on both sides of the tunnel decreased, with a more pronounced reduction on the excavation side. As the width and depth of the foundation pit increased and the horizontal net distance decreased, the absolute value of the additional surrounding soil pressure increased, and the deformation mode shifted to "lateral expansion and vertical contraction". After the excavation of the foundation pit directly above the tunnel, the overall surrounding soil pressure decreased significantly but retained a symmetrical distribution. With increasing width and depth of the foundation pit, the horizontal and vertical convergence deformation of the tunnel gradually diminished, approaching zero before transitioning to a "lateral contraction and vertical expansion" mode.

Due to the frequent accidents at the tunnel construction site, using robots instead of manual safety inspections can effectively protect the safety of staff. However, most of the robots in this field currently require manual control and lack autonomous obstacle avoidance capabilities. Aiming at the obstacle detection method in tunnel construction environment, a lightweight model using improved YOLOv5 for obstacle detection was proposed. Firstly, the obstacle dataset in the tunnel scenario was constructed. Secondly, the backbone network was modified to a lightweight Shufflenet v2 network, and the activation function was modified to the SiLU function to improve the detection speed and reduce the amount of computation. Next, the coordinate attention mechanism was incorporated to improve the network's ability to learn and represent features. Finally, the neck convolution block was modified to GSConv, which reduced the calculation amount of the model and improved the detection accuracy of the algorithm. Comparative experiments on the dataset constructed in this research showed that the detection speed of the proposed method was increased by approximately 57% compared to the original YOLOv5-n algorithm, which reduced the hardware requirements of the model.

Mining roadways were found to be susceptible to severe roof subsidence, large deformation, and structural failure under intense mining-induced disturbances. Accurately predicting deformation trends was considered crucial for ensuring mine safety and achieving effective surrounding rock control. Based on deformation monitoring data from mining faces at the Xin Julong Coal Mine, the deformation evolution characteristics of mining roadways were analyzed, and an empirical prediction model suitable for coal mining roadways was proposed. Considering that traditional empirical models were limited in capturing dynamic deformation behavior and parameter uncertainties, a Bayesian updating algorithm was introduced to construct a dynamically updated prediction model. By continuously adjusting the posterior distribution of model parameters with real-time monitoring data, prediction accuracy was improved, and uncertainty was reduced. Model validation was performed using monitoring data from working faces 6305 and 2305 of Xin Julong Coal Mine. It was indicated that, as data accumulated and Bayesian updating iterations proceeded, posterior parameter estimated stabilized, significantly enhancing prediction accuracy. The final predicted deformation values were found to closely match measured values, with determination coefficients(R2)exceeding 0.98 and root mean square errors(RMSE)significantly reduced. Additionally, extended validation was conducted using monitoring data from another coal mine's working face, further confirming the generalization capability and applicability of the proposed method under different geological conditions. The Bayesian updating prediction approach proposed in this research was demonstrated to effectively addresses the dynamic variations and parameter uncertainties in surrounding rock deformation, providing reliable data support for surrounding rock stability control in coal mine roadways.

In order to accurately identify the unfavorable geological structures in tunnels through advanced geological prediction, three typical unfavorable geological structures in tunnels, namely karst, fault fracture zone and soft interlayer, were summarized. According to the combination of media, they were classified into 2 types and 4 structural models of karst, 2 types and 4 structural models of fault fracture zone, and 1 type and 3 structural models of soft interlayer. The self-compiled Python program was used for automatic modeling to establish three types of unfavorable geological structures, namely random multi-media coupling karst voids, fault fracture zones and weak interlayers. The Python program automatically called GprMax to numerically simulate the radar characteristics of the three types of unfavorable geological bodies. The results showed that the electromagnetic waves generated by the geological radar would produce secondary pseudo-geological anomalies in water; in the model where water was mixed with other media, weak reflections would appear in the area without anomalies; the karst cave presented the characteristics of "quadratic curve type", the fracture zone presented the characteristics of "strip-shaped" and the soft interlayer presented the characteristics of "stratification fault type". Combined with the actual situation of several typical tunnels of a certain highway, the above-mentioned simulated image features were applied to actual engineering through the detection of the above-mentioned typical unfavorable geological bodies by geological radar, and the results were consistent.

In order to address the design problem of the combined support of prestressed bolts and prestressed cables in super-large span tunnels, based on the theory of the perimeter rock pressure arch, the combined effect of prestressed bolts and prestressed cables, along with the anchoring parameters, was analyzed. The support effect was evaluated in the large span transition section of Badaling Great Wall Station. It was found that the optimal arch axis of a pressure arch followed an elliptical shape, with the semi-axes were determined by the in-situ stress and the cross-sectional geometry. The axial force of the pressure arch was determined by the coincidence of the short axis direction and long axis direction of the pressure arch ellipse with the excavation section. Prestressed bolts were observed to interact with shallow surrounding rock to form pressure arches with bearing capacity, and the anchoring parameters were found to determine the shape and compressive strength of the pressure arches. Prestressed cables were utilized to fully exploit the bearing capacity of deep, stable surrounding rock, with the anchoring parameters determining their ultimate bearing capacity. The supporting effect of the combined support of prestressed bolts and prestressed cables was analyzed through the settlement of the vault. The maximum settlement of the vault was measured at 36.9 mm, indicating that the combined support system of prestressed bolts and prestressed cables effectively controls the deformation of surrounding rock.

In this research,the smoke flow characteristics of a scenic sightseeing tunnel in different fire scenarios were investigated based on numerical calculation methods. It was found that the location of the fire source had a significant effect on the flame spreading speed, and the flame spread faster to the fire elevator end when the fire point was close to the entrance of the tunnel; at the early stage of spreading, the diffusion speed of the lower surface was larger than that of the upper surface. As the intensity of the fire source increased, the ignition time of the sightseeing screen was gradually shortened from 52 s to 29 s. When the intensity of the fire source increased from 0.5 MW to 1.0 MW, the ignition time was shortened by about 20 s, and the evacuation time was greatly advanced. The research also showed that when the intensity of the fire source was less than 1.0 MW, the time to reach the critical point for safe evacuation of the tunnel due to spontaneous ignition of the electrical wiring was even shorter than that due to ignition, at 92 s, which was about 20% earlier. However, the fire hazard caused by the ignition of electrical wiring was more serious when the source of ignition exceeded 1.0 MW. Although the positive-pressure air supply in the elevator vestibule suppressed the spread of smoke to some extent, the evacuation conditions at the fire elevator end were still poor, with CO concentrations of up to 920 mg/m³, which could not meet the evacuation requirements.