Development of the thermophysical parameter tester for measuring the soil matrix suction and optimization of the calibration function

The accurate measurement of the soil matrix suction is the prerequisite for understanding the mechanism of unsaturated soil. In this study, the in-situ thermophysical parameter tester was developed for measuring the soil matrix suction. The determination of the soil–water characteristic curve (SWCC) is a key step in the determination of the calibration function. Therefore, the simulation performances of three theoretical models and three machine learning models on SWCC were compared. It was concluded that the particle swarm optimization extreme learning machine (PSO-ELM) model outperforms the other models. The calibration function was obtained by combining the PSO-ELM model with the line heat source theory, and the coefficient of determination (R2) of the calibration model was greater than 0.95. The calibration function was then applied to a field test in order to verify the performance of the designed instrument and compare it with the tensiometer. Finally, the error propagation and synthesis theory based on random error was adopted, and it was deduced that the theoretical systematic error of the instrument is 9.85 %. Considering that the tensiometer also has a test error, it is believed that the relative measurement error of the two instruments is reasonable. Therefore, the instrument designed with the optimal calibration function can effectively measure the soil matrix suction. It also allows to determine the temperature, thermal conductivity, and thermal diffusion coefficient, providing an accurate measurement method for the soil matrix suction.