Investigating the Mechanical Behaviour of Sand Mixed with Recycled Waste Material: Experimental and Machine Learning Approaches

Abstract

The escalating global waste crisis underscores the urgent need for comprehensive waste management strategies, with recycling practices at their core. Rubber tyre and carpet fibre waste materials are two of the major contributors to this human-related waste. These materials can be effectively recycled and utilised in various applications, including the geotechnical field. This thesis addresses environmental concerns by exploring the mechanical behaviour of cohesionless soil and investigating its behaviour when mixed with recycled waste materials. The mechanical behaviour examined includes shear strength characteristics, stiffness, compressibility, and peak interface friction. These aspects of mechanical behaviour will be examined under various influence parameters, including void ratio, particle size, particle shape, size ratio, and normal stresses.

An extensive experimental study was conducted using direct shear and oedometer test apparatus. A modified direct shear mould was designed/built and used for the interface shear test in this study. The shape of the granular material was determined using a microscope. The sample was prepared using either the pluvation technique or dynamic compaction. Furthermore, this study utilises an artificial intelligence modelling approach (machine learning) that recently attracted the attention of geotechnical researchers as it can model complicated soil behaviour considering all of its parameters of influence.

In this study, an analysis of both the macro and micromechanical behaviours of coarse-grained soil and sand mixed with recycled waste material was conducted, offering new insights into their properties and enhancing understanding. Generally, the findings indicate that as the mean particle size of the sand grains increases in the sand size range from fine to medium and from medium to coarse, both density and shear strength increase, while specific gravity decreases. In sand-rubber mixtures, the compressibility of rubber affects the consistency of the friction angle. As normal stress increases, the friction angle decreases. Hence, the traditional Mohr–Coulomb criterion may be unsuitable for such mixtures. In a mixture of sand and carpet fibre, the inclusion of the fibre reduces the density and increases shear strength. When sand is reinforced with carpet fibre in a loosely packed state, it provides higher shear strength than pure dense sand. In the interface shear experiment, the shear strength of the sand markedly exceeds the peak interface friction shown at both the smooth steel and rough steel interfaces. Furthermore, an increase in the sample density leads to an increase in the peak friction angle, due to an increase in surface contact. Moreover, machine learning models predict the experiment results with high accuracy, highlighting the influential parameters.