Most of the robust artificial intelligence (AI)-based constitutive models are developed with synthetic datasets generated from traditional constitutive models. Therefore, they fundamentally rely on the traditional constitutive models rather than laboratory test results. Also, their potential use within geotechnical engineering communities is limited due to the unavailability of datasets along with the model code files. In this study, the data-driven constitutive models are developed using only laboratory test databases and deep learning (DL) techniques. The laboratory database was prepared by conducting cyclic direct simple shear (CDSS) tests on reconstituted sand, that is, PDX sand. The stacked long short-term memory (LSTM) network and its variants are considered for developing the predictive models of the shear strain (gamma [%]) and excess pore pressure ratio (ru) time histories. The suitable input parameters (IPs) are selected based on the physics behind the generation of ru and gamma (%) of the liquefiable sands. The predicted responses of gamma (%) and ru agree well in most cases and are used to predict the dynamic soil properties of the PDX sand. The same modeling framework is extended for other sand and compared with existing AI-based constitutive models to verify its practical applicability. In summary, it is observed that though the trained models predicted the time histories of ru and gamma reasonably well; however, they struggled to predict the hysteresis loops at higher cycles. Therefore, more research is needed to verify and enhance the predictability of existing AI-based models in the future before using them in practice for simulating cyclic response.

The resilient modulus (MR) of different pavement materials is one of the most important input parameters for the mechanistic-empirical pavement design approach. The dynamic triaxial test is the most often used method for evaluating the MR, although it is expensive, time-consuming, and requires specialized lab facilities. The purpose of this study is to establish a new model based on Long Short-Term Memory (LSTM) networks for predicting the MR of stabilized base materials with various additives during wet-dry cycles (WDC). A laboratory dataset of 704 records has been used using input parameters, including WDC, ratio of calcium oxide to silica, alumina, and ferric oxide compound, Maximum dry density to the optimal moisture content ratio (DMR), deviator stress (sigma d), and confining stress (sigma 3). The results demonstrate that the LSTM technique is very accurate, with coefficients of determination of 0.995 and 0.980 for the training and testing datasets, respectively. The LSTM model outperforms other developed models, such as support vector regression and least squares approaches, in the literature. A sensitivity analysis study has determined that the DMR parameter is the most significant factor, while the sigma d parameter is the least significant factor in predicting the MR of the stabilized base material under WDC. Furthermore, the SHapley Additive exPlanations approach is employed to elucidate the optimal model and examine the impact of its features on the final result.

Hydromechanical behaviour of unsaturated expansive soils is complex, and current constitutive models failed to accurately reproduce it. Different from conventional modelling, this study proposes a novel physics-informed neural networks (PINN)-based model utilising long short-term memory as the baseline algorithm and incorporating a physical constraint (water retention) to modify the loss function. Firstly, a series of laboratory tests on Zaoyang expansive clay, including wetting and drying cycle tests and triaxial tests, was compiled into a dataset and subsequently fed into the PINN-based model. Subsequently, a specific equation representing the soil water retention curve (SWRC) for expansive clay was derived by accounting for the influence of the void ratio, which was integrated into the PINN-based model as a physical law. The ultimate predictions from the PINN-based model are compared with experimental data of unsaturated expansive clay with excellent agreement. This study demonstrates the capability of the proposed PINN in modelling the hydromechanical response of unsaturated soils and provides an innovative approach to establish constitutive models in the unsaturated soil mechanics field.

Cement-admixed clay (CAC) is a widely-used soil stabilization technique for enhancing the strength and stiffness of soft clay. However, the stress-strain behavior of CAC is complex and nonlinear, and also depends on various factors such as mixing proportion, confining pressure, stress path, and shearing condition. In this study, we propose a novel approach for modeling the stress-strain behavior of CAC using recurrent neural networks (RNNs), which are a type of deep learning (DL) technique that can well capture the temporal dependencies and nonlinearities in sequential data. We compare three types of RNNs: traditional RNN, long short-term memory (LSTM) neural network, and gated recurrent unit (GRU) neural network, and evaluate their performance in simulating the strain- and stress-controlled triaxial test results of 25 CAC specimens with different mixing proportions and confining pressures. The results demonstrate that the LSTM model, incorporating a 2-time step backward, exhibits superior prediction accuracy and generalization capability compared to other evaluated models, achieving a mean absolute percentage error (MAPE) of 4%. This LSTM model is capable of capturing the stress-strain behavior of CACs across various loading conditions and mixing proportions within a unified framework. Therefore, we suggest that the LSTM model is a promising tool for modeling and analyzing the mechanical behavior of CAC in geotechnical engineering applications.