青藏工程走廊冻融土热扩散系数特性与预测模型研究北大核心CSCD

热扩散系数是多年冻土对外界热扰动敏感程度的重要影响参数之一,也是寒区工程设计与建设的关键基础数据。基于瞬态平面热源法导热系数测试结果和质量加权法计算获取的比热容理论值,计算获得青藏工程走廊西大滩—唐古拉山沿线典型类别土样热扩散系数,分析对比了走廊带内冻融土热扩散系数的分布特征和参数影响规律,提出了基于经验拟合公式法和RBF神经网络方法的冻融土热扩散系数预测模型,并比较了不同预测模型的预测效果。研究结果表明:青藏工程走廊带内土的热扩散系数与粒径整体呈正相关性,融土热扩散系数按黏性土、粉土、全风化岩类、砂土及碎石土依次增大,冻土热扩散系数按黏性土、全风化岩类、粉土、碎石土及砂土依次增大;热扩散系数与容重及天然含水率相关性随土类及冻融状态差异明显,冻、融土热扩散系数呈显著正线性关系;以融土热扩散系数为拟合参数的冻土热扩散系数三元预测模型的预测精度明显高于二元经验公式;RBF神经网络模型在冻、融土热扩散系数预测中均具有最优的预测精度,为最佳预测模型。

Study on characteristic and prediction model of permafrost thermal diffusivity of Qinghai-Tibet engineering corridor

Thermal diffusivity is a prime influencing factor of permafrost thermal response sensitivity to external heat disturbances， and it is also the key basic data for engineering design and construction in cold regions. In present work， thermal diffusivity of typical soil samples along the Xidatan-Tanggula Mountain of the Qinghai-Tibet engineering corridor were calculated using the thermal conductivity test results of the transient plane heat source method and the theoretical value of specific heat calculated by mass weighting method. Then， the distribution characteristics and the parameter influence law of thermal diffusivity of frozen and unfrozen soil in the corridor zone had been analyzed and compared. Finally， the prediction models of thermal diffusivity of frozen and unfrozen soil based on empirical fitting formula method and RBF neural network method had been developed and compared. The research results showed that： （1） The thermal diffusivity of the soil in the Qinghai-Tibet engineering corridor is positively correlated with the particle size. The increasing order of thermal diffusivity value of unfrozen soil is cohesive soil， silt， fully weathered rock， sandy soil and gravel soil， and order is cohesive soil， fully weathered rock， silt， gravel soil and sandy soil for frozen soil. （2） The correlation among thermal diffusion coefficient， volume weight and natural moisture content varies with soil types and freezing/thawing states， and the thermal diffusivity of frozen and unfrozen soil has a significant positive linear relationship. （3） Using thermal diffusivity of unfrozen soil as fitting parameter， the prediction accuracy of ternary fitting prediction model is significantly higher than that of binary fitting model. （4） The RBF neural network model has the greatest prediction accuracy for both frozen and unfrozen soils， which is the best soil thermal diffusivity prediction model.