Zhisheng Xu

The integration of computational fluid dynamics (CFD) with deep learning in tunnel fire research is currently constrained by excessive reliance on manual operations and low overall efficiency. To address these limitations, this study presents a multi-agent collaborative framework driven by large language models, which enables full automation of the fire source characteristic inversion process. This framework reorganizes the conventional research pipeline into four dedicated, specialized agents: physical modeling, data governance, model training, and evaluation analysis. As a typical automated verification task, five deep learning models are systematically benchmarked under 45 experimental configurations to implement multi-task continuous regression inversion, which fully demonstrates the framework’s capability of automated, reproducible and large-scale comparative experiments. The experimental results demonstrated that the CNN-LSTM model outperforms other models in extracting spatiotemporal correlation features from temperature time-series data, enabling high-precision prediction of multiple fire parameters. With a 6 s observation window and 10 m sensor spacing, the average R² attains 0.942, an improvement of 2% over the baseline LSTM model, and the RMSE decreases by 28.8%. For sparse sensor deployment at 30 m spacing, the average R² remains at 0.917, confirming the effectiveness of integrating spatial feature extraction with temporal modeling. This study provides an efficient technical pathway for intelligent tunnel fire identification and advances the research paradigm by shifting the traditional manual optimization process to a multi-agent system-based optimization workflow.

To address the issues of manual operation dependency and low efficiency in tunnel fire research combining computational fluid dynamics (CFD)with deep learning, this paper proposes a multi-agent collaborative framework based on large language models to automate the entire process of inverting fire source characteristics. The framework decomposes the traditional workflow into four specialized agents, namely physical modeling, data governance, model training, and evaluation analysis, which collaboratively execute end-to-end tasks from CFD scenario generation to model deployment. The results demonstrate that the CNN-LSTM model performs optimally. Under a 6 second observation window and 10 meter sensor spacing, the average R² reaches 0.942, representing a 2% improvement over the baseline LSTM model, while the RMSE is reduced by 28.8%. Under sparse deployment with 30 meter spacing, the average R² remains as high as 0.917, validating the effectiveness of integrating spatial feature extraction with temporal modeling. This work provides an efficient technological pathway for intelligent tunnel fire identification and advances the research paradigm from manual optimization to multi-agent system optimization.

Petrochemical fires pose severe threats to public safety and environmental sustainability, necessitating urgent advancements in efficient and eco-friendly fire suppression technologies. This study systematically investigated the influence of gas–liquid ratio (GLR) on the foam properties and fire suppression efficacy of a novel short-chain fluorocarbon (PFH-BZ) foam fire extinguishing agent. Through comprehensive experimental analysis, the underlying mechanism governing foam performance was elucidated, and the burn-back resistance of optimized formulation was evaluated. The results indicate that GLR significantly impacts PFH-BZ foam performance. Foaming capacity and structural stability exhibit a positive correlation with increasing GLR until reaching a plateau. Low GLRs result in insufficient foam formation and thermal stability, while inducing detrimental combustion intensification. Conversely, excessively high GLRs impair foam spreading capacity, hindering rapid extinguishment. The optimal fire extinguishing performance is achieved at a GLR of 12, where PFH-BZ foam attains an excellent balance among drainage characteristics, spreading capacity, and structural stability. This optimized formulation achieves complete extinguishment within 26.16 s and maintains burn-back resistance of 662.37 s while effectively mitigating the vapor explosion phenomenon. These findings provide critical guidance for the application of a PFH-BZ-based foam extinguishing agent and deepen understanding of the influence of system parameters on suppression performance.

Addressing the environmental challenges posed by traditional foam extinguishing agents containing persistent pollutants, the development of eco-friendly alternatives has become imperative. This study investigates the effect of xanthan gum (XG) on the fire extinguishing performance of PFH-BZ foam formulated with short-chain fluorocarbon surfactant. By analyzing foam formation, drainage characteristics, and suppression process, the underlying mechanism by which XG influences foam extinguishing performance was elucidated. The results indicate that XG exerts dual effects on foam properties. While its viscosity-increasing effect improves foam stability, excessive XG addition impairs foaming and spreading capabilities, reducing fuel surface coverage and smothering efficiency. Moreover, a high concentration of XG hinders drainage behavior, which in turn inhibits the formation of spreadable aqueous films, thereby reducing cooling and extinguishing efficiency. The PFH-BZ foam with 0.02 wt.% XG exhibits excellent foaming and spreading capabilities, enabling rapid coverage of fuel surfaces. Additionally, its moderate drainage characteristics facilitate spreadable aqueous film formation, achieving efficient cooling and smothering effects. The optimized PFH-BZ foam exhibits the shortest extinction time of 35.4 s, the lowest transient temperature rise of 60.8 °C, and the highest cooling rate of 16.8 °C/s. Environmental assessments reveal that the optimized PFH-BZ foam exhibits higher biodegradability than conventional foam.

Polymeric stabilizers play a critical role in enhancing the stability and performance of firefighting foams. This study evaluated the influence of three polymeric stabilizers (xanthan gum, XG; polyacrylamide, PAM; sodium carboxymethyl cellulose, CMC-Na) on the performance of foam solutions formulated with a zwitterionic short-chain fluorocarbon surfactant. The investigation focused on three performances: interfacial properties, apparent viscosity (at a fixed rotational speed), and foam performance, employing interfacial tension analysis, viscosity measurement, dynamic foam analysis, and foam drainage testing. Results indicate that XG and CMC-Na slightly decrease interfacial activities, reducing spreading coefficients 6.34–15.78% and 0.68–6.35%, respectively. However, these polymeric stabilizers substantially increase apparent viscosity through hydrogen bond network formation, which effectively mitigates foam coarsening and drainage. When adding 0.10 wt.% XG, the foam solution exhibits a characteristic coarsening time of 724.64 s and a 25% drainage time of 1519.15 s. Conversely, PAM exhibits a concentration-dependent dual effect. When below 0.06 wt.%, PAM enhances interfacial properties and foam stability. However, at elevated concentrations, excessive PAM aggregates at interfaces and forms entangled networks that inhibit surfactant adsorption. This impairs foam formation and accelerates foam structural evolution, increasing variation in bubble size and promoting foam drainage by 8.63–57.88%. These findings provide crucial reference for applying polymeric stabilizers in short-chain fluorocarbon surfactant systems.

Previous studies on tunnel fires have primarily focused on tunnels with flat ceilings and lacked studies on tunnels with beams. The present study is predicated on a reduced-scale tunnel model with a beam structure. Through meticulous analysis of the effects of factors such as longitudinal ventilation velocity and beam dimensions, the study unveils the distribution pattern of ceiling temperatures under the longitudinal smoke exhaust mode. The findings suggest that the presence of beams can induce turbulence in the longitudinal ventilation airflow. It has been demonstrated that the magnitude of this phenomenon is directly proportional to the spacing of the beams. This results in fluctuations in the ceiling temperature rise close to the combustion zone. The smoke storage capacity of the open cavities formed between adjacent beams is significantly affected by the beam height, thereby influencing the overall temperature rise beneath the ceiling. The greater the beam height, the higher the overall ceiling temperature rise near the combustion zone, but the lower the ceiling temperature rise downstream of the fire source. A prediction model for the longitudinal decay of ceiling temperature downstream of the fire source in tunnels with beams has been obtained. This model is related to the dimensionless beam dimension.

Accidental surface surcharge will generate additional load in the stratum, which then leads to unfavorable impacts on the underlying shield tunnel. This paper proposes a probabilistic analysis method to address this problem. In this framework, an improved soil–tunnel interaction model considering the nonlinearity of the subgrade is established at first, and the Newton–Raphson iterative solution algorithm is employed to acquire tunnel responses. Then, the random field models of the initial stiffness and the ultimate reaction of the subgrade are constructed to realize the spatial variability of soil properties. Finally, with the aid of the Monte Carlo Simulation method, the probabilistic analyses on tunnel responses are performed by combining the improved soil–tunnel interaction model and the random field model of subgrade parameters. The applicability and the superiority of the improved soil–tunnel interaction model are validated by a historical case from Shanghai Metro Line 9. The results prove that the traditional linear foundation model will overestimate the bearing capacity of the subgrade, thereby leading to overly optimistic assessments of surcharge-induced tunnel responses. This shortcoming could be addressed by the improved nonlinear soil–tunnel interaction model. The influences of spatial variability of soil properties on tunnel responses are nonnegligible. The stronger the uncertainties of subgrade parameters, in terms of the initial stiffness and the ultimate reaction concerned in this work, the higher the failure risk of the shield tunnel subjected to the surcharge. The failure modes of the tunnel subjected to the surcharge are controlled by the longitudinal curvature radius of the tunnel within the current assessment criteria, which means if this evaluation indicator can be restricted within the allowable value, then the opening of the circumferential joint and the longitudinal settlement can also meet the requirements. Compared with the influences of the uncertainty of the subgrade ultimate reaction, the spatial variability of the subgrade initial stiffness has greater influences on tunnel failure risk under the same conditions. An increase in the range of surcharge will raise the risk of tunnel failure, while the influence of tunnel burial depth is just the opposite.

Dou-Gong brackets, the distinctive structural element in ancient Chinese architecture, fulfill critical roles in load transfer, span reduction, and decoration, making its preservation vital for safeguarding wooden heritage buildings. This study investigates the combustion performance and residual load-bearing capacity of key Dou-Gong bracket components—Zuo-dou, Zheng-xin-gua-gong, and Qiao—exposed to varying fire conditions. The results reveal that an increasing fire load elevates heating rates and peak temperatures of wood substrates, resulting in a significant degradation of structural integrity. At a fire load of 55 kW, the peak temperatures at the bottom, joint edge, and top of the Dou-Gong brackets reach 755.3 °C, 489.9 °C, and 620.7 °C, respectively, representing increases of 2%, 65%, and 38%, respectively, compared to those observed at a fire load of 20 kW. Moreover, the charring rate of Dou-Gong bracket increases from 0.22–0.26 mm/min at a fire load of 20 kW to 0.50–0.56 mm/min at a fire load of 55 kW, accompanied by an increase in mass loss rate from 28.5% to 36.9%. These findings highlight the significant impact of fire conditions on the fire characteristic and structural integrity of Dou-Gong brackets, providing the first quantitative evidence of their degradation under fire exposure. By addressing this vulnerability, the study contributes to the scientific preservation of ancient wooden architecture under contemporary fire risk scenarios.