Liquefied petroleum gas (LPG) leaking in underground spaces can easily form explosive mixtures; traditional ventilation methods have limited effectiveness, and existing research on non-thermal plasma (NTP) has primarily focused on low-concentration VOCs, making it difficult to apply to scenarios involving high-concentration LPG leaks. In this study, C4H10 was used as a model for LPG. A NTP reactor was employed to investigate the effects of discharge power, gas flow rate, and initial C4H10 concentration on reaction performance through comprehensive experiments and multiple regression analysis, and a quasi-first-order kinetic model of the reaction was established. The results indicate that the initial C4H10 concentration is the primary factor affecting the conversion rate and the residual combustible concentration in the exhaust gas, while gas flow rate has the most significant impact on energy yield, liquid product selectivity, and CO by-product formation, the secondary regression models constructed all had coefficients of determination above 0.90; A multiple regression model used to optimize the operating parameters showed that the C4H10 mixture with an initial concentration of 13.32 % can be reduced to the lower explosive limit of 1.9 %; For a C4H10 leak scenario with an concentration of 5 %, the optimal control can limit the C4H10 concentration in the exhaust gas to 2.47 %–2.66 % and CO to 0.4 %–0.6 %, thereby balancing the risks of explosion and CO poisoning; NTP partial oxidation of C4H10 in an air atmosphere follows quasi-first-order kinetics, with the reaction rate constant linearly correlated with discharge power; the maximum absolute error in the model’s conversion rate prediction is only 3.6 %. This study provides process guidance and theoretical support for active explosion prevention measures in response to high-concentration LPG leaks in confined spaces.
Deep-stratum non-coal gas reservoirs are characterized by low permeability, high in-situ stress, and limited gas reserves. The gas migration and release laws of surrounding rocks in such strata under excavation disturbance have not been systematically investigated. Based on field cases of gas emission in magmatic rock strata, this study explores the formation and occurrence mechanisms of high-pressure gas chambers. According to the unique characteristics of high-pressure gas reserves in magmatic rocks, a physical model test system for gas migration and release during tunnel excavation is developed. The stress response laws of surrounding rocks under the coupling effect of gas pressure and excavation unloading are revealed, and four typical evolutionary stages of gas emission from high-pressure gas chambers in magmatic rocks are proposed. Considering the significant fluctuation characteristics of pressure boundaries in small-scale gas reserves, a dynamic gas pressure boundary calculation method is established. A steady-state damage and transient seepage iterative coupling approach is adopted to realize the full-process evolutionary calculation of surrounding rock damage, deformation, and gas seepage. The dominant factors governing gas emission at different excavation stages are clarified, and a prediction model for gas emission volume of high-pressure gas chambers is established with comprehensive consideration of fracture density and in-situ stress. The research findings provide theoretical support for the prediction and control of gas outburst disasters in tunnels excavated in non-coal strata.
Accurate detection and quantitative characterization of rock joint traces are fundamental for rock mass stability assessment in underground engineering. However, automated analysis remains limited by the scarcity of labeled tunnel images and by the poor suitability of orientation-agnostic detectors for thin, elongated joint traces. We propose a cascaded framework integrating a diffusion-based joint-trace image generation with an Orthogonal Decomposition Joint-Trace-aligned Adaptive Sampling Scheme (OD-JTAS) in this study. The generative module synthesizes label-paired tunnel images by decoupling skeleton morphology from rock-wall texture, thereby expanding the training distribution without additional manual annotation. OD-JTAS further restructures deformable attention sampling into longitudinal and transverse components aligned with the dominant joint orientation. The resulting instance masks are converted into three engineering indicators: areal fracture intensity (P21), box-counting fractal dimension, and fracture-network topology.Validation using tunnel images from the Jinchuan Mining Area, Gansu Province (China), demonstrates that diffusion-generated data increase mean Average Precision at an IoU threshold of 0.50 (mAP@0.50) from 0.5874 to 0.7528, where the complete OD-JTAS model achieves the best detection performance among all the tested methods. The framework therefore provides a unified image-based route from joint detection to quantitative rock mass characterization.
Excavation-face instability during deep tunnelling can trigger progressive soil collapse and induce simultaneous deformation of adjacent structures. Existing studies have mainly focused on face stability or the response of individual tunnels. However, it remains unclear how excavation configuration alters soil-collapse propagation and influences the deformation of multiple tunnels within a dense underground complex. To address this gap, a series of full-scale three-dimensional coupled finite element method–smoothed particle hydrodynamics models are established based on a previously validated numerical framework. Excavation-face instability is represented by instantaneous complete loss of face support. The excavation configuration is varied by changing the depth, diameter, and horizontal offset of the excavation face one at a time to examine the resulting changes in soil-collapse propagation, tunnel deformation, soil–tunnel contact pressure, additional internal forces, and performance degradation. A deformation-based assessment framework is further employed to quantify segment-level residual performance and the overall failure degree of the evaluated tunnel system. The results show that the existing structures redirect the propagation of collapsed soil and cause localized deformation and load redistribution. Within the investigated configurations, increasing the excavation diameter is accompanied by a larger absolute disturbed soil volume, more pronounced tunnel deformation and internal-force redistribution, and greater performance degradation, while the effects of excavation depth vary among response measures and tunnels. Changing the horizontal offset primarily shifts the disturbed region and the affected segment ranges.
The transition from distributed microcracking to localized fracture in steel-fiber-reinforced concrete (SFRC) segmental linings is difficult to identify because visible cracking, internal fracture activity, and structural deformation may evolve asynchronously. A full-scale rebar-free SFRC segmental lining was tested under staged loading and differential lateral unloading. Mechanical measurements, machine-vision crack observations, acoustic emission (AE) and microseismic (MS) monitoring, and Gaussian mixture model (GMM) analysis of AE feature populations were synchronized using horizontal convergence. A system-normalized AE-ratio was introduced to track the relative accumulation of AE and MS responses. The ratio remained at approximately 0.85–0.90 during the pre-localization stage and decreased rapidly over a dominance-transfer interval of 77–91 mm horizontal convergence. This interval occurred mainly during late differential lateral unloading, as the generalized load difference increased from 151.4 to 155.1 kN, the opening of the B2–L1 joint increased from approximately 9.5 to 11 mm, and existing cracks continued to widen. Segment B3 exhibited a segment-body-dominated path with persistent mixed RA–AF feature populations, whereas segments B2 and L1 showed discontinuity-controlled paths characterized by late-stage enrichment of low-RA AE features and crack growth near joints. Crack-width growth, AE feature-population reorganization, and spatial concentration of AE/MS events around the final crack zones showed that the response transfer coincided with structural localization. The procedure provides a deformation-synchronized means of relating AE/MS response evolution to progressive localization at ring and segment scales.
