Improved Prediction of Frost Depth Penetration Using Recurrent Neural Networks

Frost effects, such as frost heave and thaw weakening, can significantly degrade pavements. Effective mitigation relies on accurate prediction of frost penetration depth. Current prediction methods use empirical equations or finite element analysis, with recent progress using machine learning. One potentially appropriate machine learning model may be a recurrent neural network, which takes in present data as input and outputs an estimation of future data, which can then be fed back into the model recursively to make further predictions. Using this method and training data from Hill Air Force Base, Utah, and Air Force Academy, Colorado, we were able to forecast soil parameters including temperature, thermal conductivity, and moisture content for frost susceptible soils, with a deviation from experimental values of no more than 10%, with the most significant contributions to accuracy being the use of gated recurrent unit neurons and the incorporation of multiple soil parameters.