The thermal conductivity characteristics and prediction models of limestone sandyellow soil mixtures

To optimize the backfilling of ground source heat pump drilling mud and boost the thermal conductivity of drilling materials, this study proposes using a mixture of limestone sand and loess, typical in karst regions, as backfill for buried pipe heat exchangers. Through indoor experiments, 152 limestone sand-loess mixtures were prepared and their thermal conductivities tested. Analyses explored the impacts of limestone sand content, moisture content, dry density, and particle size distribution. Results show that artificially graded materials generally outperform natural ones in thermal conductivity, with grading's influence decreasing as moisture rises. At 8% moisture, grading increases thermal conductivity by 18.57% (0.069 - 0.124 W/(m·K)); at 20%, the increase is 7.63%. High moisture and limestone sand content can yield a thermal conductivity of 1.508 W/(m·K). When using graded materials, geological conditions and aquifers should be considered, and they suit strata with moderate moisture. A backpropagation neural network - based predictive model for thermal conductivity, developed from experimental data, achieved 6.4% average absolute percentage error, indicating good accuracy.