Improvement of normalized prediction model of soil thermal conductivity

Abstract

Soil thermal conductivity (k) is an important parameter for geotechnical engineering temperature field analysis and energy pile design. It is affected by various factors such as degree of saturation, dry density, and mineral composition. Due to the nonlinear correlation between k and these influencing factors, accurate prediction of k is of great significance to guide practical engineering. In this paper 337 soil thermal conductivity data were collected and the intrinsic correlations between k and the influencing factors were systematically analyzed. Based on the Johansen model, a new predictive model was developed by the multivariate probability distribution (MPD) method. The proposed model was also verified by making a comparative analysis with the empirical ones. The results show that the Johansen model has a superior performance in predicting k after the improvement of MPD analysis, in which the correlation coefficient R2 increases from 0.641 to 0.961. The prediction accuracy of the improved model is significantly better than the traditional empirical relationship models. It is recommended to use the Ke - {n, γd, Sr, Cq, Cc, Csi, Csa} model to predict the k when the soil parameters are clear.