A new method to model the thermal conductivity of soil-rock media in cold regions: An example from permafrost regions tunnel

- A multiple regression model is established to calculate thermal conductivity of soil-rock media.
- Significant testing shows that the regression effect is well.
- Compared with the other three general models, the multi-regression model is reasonable and effective.
- A general regression model for soil-rock media is given based on apparent heat capacity theory.

Thermal conductivity of soil-rock media is an important index for analyzing the temperature field of cold region engineering. As a kind of mixed media, the thermal conductivity is determined by many factors, e.g. porosity, dry density, saturation degree, temperature, water content and so on. In this study, the thermal conductivity model of soil-rock media is established by multiple linear regression method based on the basic parameters of soil-rock samples. The significant testing (correlation coefficient R2, F-test and probability-of-F-to-enter) shows that the regression effect is well. To validate the multiple linear regression model, a comparison of the calculated values from the model with the tested data not used in the construction of the model shows that this model can effectively calculate thermal conductivity of soil-rock media. Additionally, compared with the other three general models, i.e. BP neural model, physical model and CK model, the multiple linear regression model is more reasonable and effective. Finally, based on the apparent heat capacity theory, a general regression model for soil-rock media in cold regions is obtained. It is hoped that a parameter basis can be provided for the temperature field simulation of cold region engineering from the study.