Case Studies in Thermal Engineering (Sep 2024)
Evaluation of 17 thermal conductivity models for frozen soil
Abstract
Accurate prediction of thermal conductivity (λ) in frozen soils is essential for understanding their thermal behavior and improving thermal response analyses. Currently, there are numerous models for calculating λ, but unified laboratory data to compare and evaluate these models are lacking. This study focuses on evaluating the performance of 17 models for applied to frozen soils using a unified laboratorial dataset. The heat-pulse-probe (HPP) method was employed to measure the thermal conductivity of silty clay, sand, and sand loam. Nuclear magnetic resonance experiments were conducted to measure the unfrozen water content (θw) and ice content(θi). This provides unified laboratorial dataset of λ-θi-θw. The predictive performance of 17 models demonstrated that the model of Zhang et al. (2018), Sass et al. (1971) and Tian et al. (2016) are the three optimal models: 1). The model of Zhang et al. (2018) is the provided the best predictive performance, with mean absolute error (MAE) values of 0.17 Wm-1K-1, 0.07 Wm-1K-1, and 0.06 Wm-1K-1 for silty clay, sand and sand loam, respectively. This model is advantageous due to its simplicity and the absence of undetermined parameters. 2). The model of Sass et al. (1971) is suitable for soils with low sensitivity to frozen volume deformation, such as sand and sand loam, achieving an MAE of 0.06 Wm-1K-1. 3). The model of Tian et al. (2016) exhibited strong performance for silty clay with an MAE of 0.11 Wm-1K-1, contingent on the precise determination of soil-specific parameters. This study provides a unified laboratory data for the establishment of thermal conductivity models. The evaluation of existing models can serve as a reference for future related research.
