Upgrading the prediction of jet grouting column diameter using deep learning with an emphasis on high energies

Abstract

This article proposed a new method to estimate the diameter of jet grouting columns. The method uses the largest data collection of column diameters measured to date and includes a large amount of new data that fills the existing gap of data for high injection energies. The dataset was analysed using a deep neural network that took into account the problem’s key parameters (i.e. type of soil, soil resistance, type of jet and specific energy in the nozzle). As a result, three different neural networks were selected, one for each type of jet, according to the errors and consistency associated with each. Finally, using the trained networks, a number of design charts were developed to determine the diameter of a jet grouting column as a function of the soil properties and the jet system. These charts allow generating an optimal jet grouting design, improving the prediction of the diameter of jet columns especially in the high energy triple fluid.