Natural gas transportation heavily relies on buried natural gas network. Unfortunately, natural gas pipeline leakage can cause environmental pollution and resource loss, and even leads to serious risks on human's life and properties due to its flammable and explosive properties. In this study, numerical stimulation methodology was applied to investigate the leakage and diffusion characteristics of buried natural gas pipeline, and the first dangerous time (FDT), farthest dangerous range (FDR) and ground dangerous range (GDR) were utilized to track harmful boundary for gas leakage and diffusion. By exploring thirteen different working conditions and evaluating the impacts of six parameters on harmful boundary, prediction models for harmful boundary was successfully established. The results demonstrated that the effect of soil on the pressure and velocity of the leaking hole occurred within 30 s right after pipeline leaked. Sonic flow was absent at the leaking hole, while reversal flow was present at the pipeline downstream of the leaking hole. Soil types play a critical role on harmful boundary: FDT increased by more than five times, FDR decreased by more than two times and FDR remained zero when soil type switched from sand to clay. Larger leakage diameters and higher pipeline pressures can raise FDR and GDR, but reduce FDT. In this prediction models, the average errors for FDT, FDR and GDR were 8.81%, 1.51%and 3.82% respectively. This is the first study to assess harmful boundary by utilizing prediction models. This work will provide reasonable time estimation for emergency evacuation, but also suggest the location setting of underground gas leakage monitoring points.

天然气运输严重依赖埋地天然气管网。不幸的是，天然气管道泄漏会造成环境污染和资源损失，甚至由于其易燃易爆的特性对人类的生命和财产造成严重威胁。本研究应用数值模拟方法研究埋地天然气管道的泄漏和扩散特性，并利用首次危险时间（FDT）、最远危险范围（FDR）和地面危险范围（GDR）跟踪有害边界用于气体泄漏和扩散。通过探索13种不同工况，评估6个参数对有害边界的影响，成功建立有害边界预测模型。结果表明，土壤对泄漏孔压力和速度的影响发生在管道泄漏后 30 s 内。泄漏孔处无声波流动，而泄漏孔下游管道存在逆流。土壤类型对有害边界起着关键作用：当土壤类型从沙子转向粘土时，FDT 增加了 5 倍以上，FDR 减少了 2 倍以上，FDR 保持为零。更大的泄漏直径和更高的管道压力会提高 FDR 和 GDR，但会降低 FDT。在该预测模型中，FDT、FDR 和 GDR 的平均误差分别为 8.81%、1.51% 和 3.82%。这是第一项利用预测模型评估有害边界的研究。这项工作将为紧急疏散提供合理的时间估算，