Physics-Informed Ensemble Machine Learning Framework for Improved Prediction of Tunneling-Induced Short- and Long-Term Ground Settlement

Abstract

Machine learning (ML), one of the AI techniques, has been used in geotechnical engineering for over three decades, resulting in more than 600 peer-reviewed papers. However, AI applications in geotechnical engineering are significantly lagging compared with other fields. One of the reasons for the lagging is that hyperparameters used in many AI techniques need physical meaning in geotechnical applications. This paper focuses on widening the applications of ML in predicting tunneling-induced short- and long-term ground settlement and optimizing ML architectures considering their interpretability and ability to provide physically consistent results. Informed by the underlying physics knowledge, tunneling-induced ground settlement is divided into long-term and short-term settlements since different mechanisms and influencing parameters contribute to these two deformation types. Based on the above considerations, this paper introduces a physics-informed ensemble machine learning (PIML) framework to strengthen the connection between ML techniques and physics theories, followed by identifying/utilizing different sets of parameters for effectively predicting short- and long-term tunneling-induced settlements, respectively. Together with in situ observations and experimental lab results, parameters obtained from physics equations are set as inputs for the ML models. Results show that the proposed PIML framework effectively predicts tunneling-induced ground movements, with a predicting accuracy above 0.8. Additionally, parametric studies of variable significance and comparisons among different ML designs reveal that in situ observed dynamic parameters, for instance tunnel face and monitoring points (DTM), gap parameter, and tunnel depth, are essential in predicting tunneling-induced short-term settlement, while predicting long-term settlements largely depends on features, such as tunnel depth, volume compressibility, and excess pore pressure, derived from physics theories.

Evaluation of Jacking Forces in Weathered Phyllite Based on In Situ Pressuremeter Testing and Deep Learning

“…Increase in face pressure force [15,[17][18][19] Although the influences of these operation parameters are identified, the quantitative measurements of these operation parameters on jacking forces remain absent. Recent research has shown the use of machine learning and deep learning techniques in the pipe jacking process, such as predicting the changes in geological conditions [26][27][28][29], changes in ground settlement [30][31][32][33][34][35] and prediction of various operation parameters [36][37][38][39][40]. Hence, this paper will use deep learning techniques, such as gated recurrent units (GRUs) with an attention mechanism, to predict jacking forces through a region of weathered phyllite based on pipe jacking operation parameters as the input features.…”