Microbially induced carbonate precipitation (MICP) is an emerging technique for enhancing the mechanical properties of granular soils. Although several experimental studies have reported increased shear strength in MICP-treated soils at both peak and residual states, other findings have shown reductions in residual strength compared to untreated soils. This study uses the discrete element method (DEM) to investigate the mechanisms governing the residual strength of bio-cemented sands. The results indicate that residual strength may decrease when carbonate precipitates in the form of grain-bridging patterns. In that case, the introduction of carbonates alters the contact network and may induce metastable configurations, particularly when the bonds are weak or non-cohesive. These configurations are prone to strain localisation upon shearing, leading to the development of shear bands and a reduction in residual strength. Conversely, higher cohesive strength enhances microstructural stability, offsetting the weakening effects of localisation. The residual strength of bio-cemented sands is therefore governed by two competing mechanisms, namely bond-induced stabilisation and instability-driven localisation.

This study introduces a small-strain stiffness model for bio-cemented sands, building upon the existing small-strain stiffness model for sands proposed by Wichtmann and Triantafyllidis. The small-strain stiffness of numerical specimens, including sand specimens with different void ratios and bio-cemented specimens with different microscopic features, is evaluated using the discrete element method (DEM). The acquired DEM results are utilised to develop the small-strain stiffness model for bio-cemented sands. The proposed model is able to describe the small-strain stiffness of DEM bio-cemented sands. In particular, different contributions of carbonates in different distribution patterns to G0 enhancement can also be described by the proposed model.

Microbially induced carbonate precipitation (MICP) has emerged as a promising ground improvement technique, with MICP-treated soils exhibiting substantial enhancements in strength. However, experimental results revealed significant variability in strength outcomes of MICP-treated soils, even under identical treatment conditions and soil properties. This uncertainty in strength is challenging to capture using traditional predictive approaches such as conventional constitutive models. The present study leverages artificial intelligence to address the challenge by developing a Bayesian neural network (BNN) model for predicting the strength of bio-cemented soils while considering uncertainty. A dataset comprising 480 experimental samples was used to develop the model. The results indicate that carbonate content and confining pressure emerge as the most influential factors governing the strength of bio-cemented soils. The BNN model exhibits lower uncertainty when predicting bio-cemented soils with relatively low strength, while demonstrating higher uncertainty for soils with strength exceeding 2 MPa. Moreover, micromechanical investigations using the discrete element method (DEM) reveal that multiscale factors, including crystal distribution patterns, fabric and spatial heterogeneity of precipitates, contribute significantly to the strength uncertainty of bio-cemented soils. The developed BNN model provides an alternative tool for predicting bio-cemented soil strength with quantified reliability, facilitating the design of MICP treatment and its application in geotechnical engineering.

Bio-mediated methods, such as microbially induced carbonate precipitation, are promising techniques for soil stabilisation. However, uncertainty about the spatial distribution of the minerals formed and the mechanical improvements impedes bio-mediated methods from being translated widely into practice. To bolster confidence in bio-treatment, non-destructive characterisation is desired. Seismic methods offer the possibility to monitor the effectiveness and mechanical efficiency of bio-treatment both in the laboratory and in the field. To aid the interpretation of shear wave velocity measurements, this study uses the discrete element method to examine the small-strain stiffness of bio-cemented sands. Bio-cemented specimens with different characteristics, including properties of the host sand (void ratio, uniformity of particle size distribution) and properties of the precipitated minerals (distribution pattern, content, Young’s modulus), are modelled and subjected to static probing. The mechanisms affecting the small-strain properties of cemented soils are investigated from microscopic observations. The results identify two mechanisms controlling the mechanical reinforcement associated with bio-cementation, namely the number of effective bonds and the ability of a single bond to improve stiffness. The results show that the dominant mechanism varies with the properties of the host sand. These results support the use of seismic measurements to assess the mechanical efficiency and effectiveness of bio-mediated treatment.

Bio-mediated methods, such as microbially induced carbonate precipitation (MICP) and enzyme induced carbonate precipitation (EICP), have gained considerable attention as alternatives to invasive ground improvement techniques. MICP and EICP use biogeochemical processes to drive carbonate precipitation and bind soil grains, thereby improving the mechanical performance of soils. While MICP and EICP have the potential to transform geotechnical engineering, challenges persist, particularly in understanding the spatial distribution of the precipitated minerals and predicting the associated mechanical improvement. This thesis addresses these uncertainties and sheds light on the role of the microstructure on the mechanical behaviour of bio-cemented sands and the effectiveness of the bio-cementation treatment.

To this end, experimental observations on the microstructure of bio-cemented sands are first reviewed. Four typical carbonate distribution patterns are identified depending on the location of the precipitated carbonates with respect to the granular assembly: grain bridging, contact cementing, grain coating, and pore filling. The discrete element method (DEM) is then used to investigate the effect of the aforementioned carbonate distribution patterns on the mechanical behaviour of bio-cemented sands. A toolbox based on the open-source DEM platform YADE, called Cementor, is developed for modelling various crystal distribution patterns and contents.

DEM models of bio-cemented sand samples with different properties, i.e. distribution pattern and mass content of carbonates (up to 3\%), are subjected to drained triaxial compression tests under various confinements. It is found that carbonates in the pattern of coating and pore filling have a negligible influence on the mechanical response of the material. In contrast, grain bridging and contact cementing lead to a noticeable improvement in stiffness, peak strength and dilatancy. The difference in the macroscopic behaviour of the cemented samples is explained by microscopic indicators such as the effective coordination number and bond breakage evolution.

Some experimental studies surprisingly show that MICP-treated sands can exhibit a lower residual strength than uncemented sands. Such a phenomenon has been observed experimentally but not explained. The presented DEM model offers insights into this behaviour. It was found that carbonates precipitated in the bridging distribution pattern are more likely to form a metastable structure. This metastable structure is prone to relative movement of particles, which in turn leads to the development of shear bands and, overall, a lower residual strength than the uncemented sample.

The variation in the microscopic properties of bio-cemented sands leads to uncertainty of mechanical improvements gained from bio-mediated treatment. It is crucial to develop a probing method to gain confidence in the treatment. Seismic measurements can be used to probe the treated soil. To evaluate the seismic response of bio-cemented sands, DEM samples with different characteristics, including properties of the host sand (void ratio, uniformity of particle size distribution) and properties of the precipitated minerals (distribution pattern, content, Young’s modulus), are modelled and subjected to static probing to examine the small-strain stiffness. These factors are all found to affect the small-strain stiffness of bio-cemented sands, hence, the seismic response. Microscopic analysis indicates that there are two mechanisms which, together, determine the overall efficiency in improving the small-strain stiffness of bio-cemented sands: the number of effective bonds and the ability of a single bond to improve stiffness.

This thesis contributes to the understanding of the macroscopic mechanical behaviour of bio-cemented sands from the microscopic point of view. In particular, the role of crystal distribution patterns is highlighted by explicitly modelling the precipitated carbonates in pre-defined locations. The findings of this thesis can support the prediction of the mechanical behaviour of bio-cemented soils and guide the design of MICP/ EICP treatment.

Bio-cemented soils can exhibit various types of microstructure depending on the relative position of the carbonate crystals with respect to the host granular skeleton. Different microstructures can have different effects on the mechanical and hydraulic responses of the material, hence it is important to develop the capacity to model these microstructures. The discrete element method (DEM) is a powerful numerical method for studying the mechanical behaviour of granular materials considering grain-scale features. This paper presents a toolbox that can be used to generate 3D DEM samples of bio-cemented soils with specific microstructures. It provides the flexibility of modelling bio-cemented soils with precipitates in the form of contact cementing, grain bridging and coating, and combinations of these distribution patterns. The algorithm is described in detail in this paper, and the impact of the precipitated carbonates on the soil microstructure is evaluated. The results indicate that carbonates precipitated in different distribution patterns affect the soil microstructure differently, suggesting the importance of modelling the microstructure of bio-cemented soils.

Microbially induced carbonate precipitation (MICP) involves bacteria to drive calcite precipitation and naturally cement soils, thereby improving soils performance. Experimental studies have shown that bio-cemented specimen can suffer from severe spatial inhomogeneity of the calcite content, leading to large uncertainty in treatment efficiency prediction. To evaluate the effect of inhomogeneity on the mechanical behaviour of bio-cemented soils, the discrete element method (DEM) is used to model bio-cemented samples with a single carbonate distribution pattern (i.e. either bridging or contact cementing) but different characteristics of inhomogeneity. Both drained triaxial compression and triaxial extension simulations are carried out to evaluate the impact of inhomogeneity along different loading paths. The results indicate that inhomogeneity has different effects on bio-cemented samples depending on the carbonate distribution patterns and the loading path. Specifically, the shear strength in compression of samples exhibiting bridging cementation is largely affected by inhomogeneity, while the effect on shear strength in extension is negligible. On the other hand, samples with contact cementing show limited sensitivity to the variation of inhomogeneity under both triaxial compression and triaxial extension tests.

Microbially induced carbonate precipitation is a promising ground improvement technique which can enhance the mechanical properties of soils through the precipitation of calcium carbonate. Experimental evidences indicate that the precipitated carbonate can display different distribution patterns. Crystals can develop at grain–grain contacts (contact cementing), connect soil grains that were initially not in contact with each other (bridging), precipitate on the grain surface (coating), or fill in the void space (pore filling). This paper investigates the role of the aforementioned distribution patterns on the mechanical behaviour of lightly bio-cemented soil samples using discrete element modelling. Bio-cemented samples with different distribution patterns and carbonate contents are built, and a series of drained triaxial compression simulations are carried out at different confining pressures. The results show that cementation in the form of bridging and contact cementing leads to obvious improvement in stiffness, strength and dilatancy. In contrast, cementation in the form of coating contributes only slightly to mechanical improvement, and pore filling exhibits negligible influence on the mechanical response of the material. The findings suggest that, to gain strength improvement in the most effective way, treatments should be tailored to precipitate calcium carbonate crystals in the form of bridging.