Double-layered granular soil modulus extraction for intelligent compaction using extended support vector machine learning considering soil-structure interaction

Abstract

Intelligent Compaction (IC) has been acquiring a growing interest in real-time quality control of compacted soil layers because of its high efficiency and full-area coverage. The current intelligent compaction technology allows the determination of the uniformity level of compaction over large areas according to the dynamic response of the roller. However, accurate real-time determination of the soil modulus during compaction based on roller acceleration has been challenging due to the multi-layered composite nature of the soil and the nonlinearities of the governing dynamic equations of motion and soil response. This study adopts a double-layered soil profile, and a three-dimensional finite element model, accounting for soil-drum interaction, is utilised for the analysis. The isotropic hardening elastoplastic hysteretic model was implemented to simulate the soil behaviour subjected to cyclic loading ranging from small to large strain amplitudes and account for stiffness degradation. The comprehensive dataset composed of the roller acceleration response and ground characteristics is then used to correlate the predicted soil modulus via an advanced machine learning approach. The adopted machine learning method incorporating Gaussian Kernel and Generalised Gegenbauer Kernel functions can reasonably predict the double-layered soil modulus during roller compaction. Additional analyses were conducted to observe the proper training size and number of iterations to achieve real-time quality control to be used by site engineers. Furthermore, the influences of the relative modulus ratio, drum length and top layer modulus on the soil surface dynamic displacement are discussed.