In order to accurately predict the thermal conductivity of unsaturated frozen soil

the characteristic structure identification method and reconstruction method of unsaturated frozen soil were proposed based on the soil microstructure images

and the prediction model of the thermal conductivity of unsaturated frozen soil was established by combining these methods and the conventional finite element method. Through scanning electron microscope (SEM) images

the content

size and distribution probability of each component were identified by antidromic quartet structure generation set (AQSGS) method. A multi-element quartet structure generation set method considering soil

water

ice and gas (MQSGS method) was proposed to improve the conventional quartet structure generation set (QSGS) method. Based on the established unsaturated frozen soil model

the thermal conductivity of unsaturated frozen soil was obtained through Monte Carlo method

and compared with the thermal conductivity of frozen soil in the specification

which verified the rationality of the prediction model (average error <4%). The influences of porosity

particle size

soil particle thermal conductivity

degree of saturation and freezing rate on the thermal conductivity of unsaturated frozen soil were studied by multi-factor analysis. The correlation coefficients between each influencing factor and thermal conductivity were −0.352

−0.098

0.641

0.52 and 0.06

respectively. The influence sequences were soil particle thermal conductivity > degree of saturation > porosity > soil particle size > freezing rate. The effects of various influencing factors on the thermal conductivity of unsaturated frozen soil can be summarized as the influences on the density

width

connectivity

heat flow capacity of "thermal chain" formed by heat flux and "thermal bridge" flux.