Thermophysical Parameter Estimation of Lunar Regolith Simulant Using Robotic Thermal Probe and Data-Driven Analytics

The measurement of subsurface thermophysical parameters is an important part of extraterrestrial exploration. Because the contact thermal resistance cannot be accurately estimated in thermal problems, it is often difficult to have an accurate solution for thermal probe models. In this work, we develop a robotic thermal probe based on the transient hot wire method to measure the thermophysical parameters of the lunar soil, and a data-driven analytic method to interpret the temperature data of the formation through the active heating. This device is conveyed by a drill pipe and penetrated into the sidewall of borehole by a piercing mechanism. In this way, the contact thermal resistance is minimized in the regolith. A finite element model is built to create the training dataset. In addition, a vacuum vessel is built to simulate the probe under the lunar environment. An inversion model based on enhanced machine learning algorithm with Newton iterative method is proposed to obtain the accurate prediction of thermal physical parameters. Especially, for the lunar regolith simulant with a low thermal conductivity, the accuracy of this improved neural network model is more than 92%. Overall, it provides a novel technique to measure the thermophysical parameters of regolith on the moon.