Machine Learning–Finite Element Mesh Optimization-Based Modeling and Prediction of Excavation-Induced Shield Tunnel Ground Settlement

Analysis of ground deformation development and settlement prediction by air-boosted vacuum preloading

Ground settlement during shield tunnel boring is a construction problem closely related to a variety of factors. To better control ground settlement, this paper proposes a ground settlement control method based on three collaborative methods: ground settlement prediction by FLAC3D numerical simulation, ground penetrating radar (GPR) backfilling grouting quality inspection and surface settlement monitoring. The results show that: the settlement pattern calculated by numerical simulation during excavation satisfies the peck settlement curve, and the maximum settlement does not exceed 6mm; the quality monitoring and dynamic feedback mechanism of backfill grouting can effectively guide the grouting construction and ensure compact grouting, and the results of monitoring the settlement of the tunnel surface show that the settlement of the surface and the settlement and tilt values of the building are within the permissible range.

In this study, a method for calculating the ground loss-induced ground settlement during quasi-rectangular shield (QRS) tunnelling was investigated. Based on stochastic medium theory, an equation for calculating QRS tunnelling-induced transverse ground settlement was derived. A tunnelling face convergence mode coefficient, α, was introduced to make the derived equation applicable to various tunnelling face convergence modes. A modified equation for calculating the longitudinal variation in the ground loss ratio (e) was proposed to calculate e for any arbitrary cross-section in the tunnelling direction. On this basis, the proposed equation was longitudinally expanded, and an equation for calculating the three-dimensional (3D) ground settlement was obtained. A simplified model for additional grouting pressure was established, and the influence of additional grouting pressure on surface deformation was analysed. In addition, the proposed method was used to calculate the QRS tunnelling-induced ground settlement, and the results were compared with actual measurements. It is found that the predicted ground settlement obtained using the proposed method matches the actual measurements relatively well and reflects the 3D variations in the ground settlement. As α increases, the maximum ground settlement increases and the width of the ground settlement trough decreases, whereas the positions of the inflection points of the settlement curve remain unchanged. The horizontal distance between the inflection points is approximately twice the width of the tunnel.

There is much uncertainty to estimate the ground settlement associated with shield tunnelling due to the complicatedly interacted conditions between tunnelling methods and soil characteristics. Accordingly, if the shape of the ground settlement could be quantitatively predicted, it would be very effective for the assessment of its influence on the existing structures. It is presented that the shape of the ground settlement trough can be expressed as the product of the normal distribution function and the error function by introducing the mathematical model. The field observations were also carried out at the four sites of the shield tunnel construction in the Metropolitan areas. Besides, the model test was done, in which the settlements of sand layers were measured by the X-ray techniques. The following conclusions are obtained. The lateral and longitudinal profiles of the ground settlement trough are well fitted with the normal distribution curve and the error function curve, respectively. The parameter which represents the width of settlement is roughly proportional to depth of tunnel and its proportional constant depends upon soil characteristics. In clay the ratio of lost ground is positively correlated with OFS (Simple Overload Factor) defined by construction conditions and soil characteristics. Consequently the maximum settlement can be calculated by using these relationships.

최근 급격한 도시화로 인한 신도시, 택지개발지구 등의 증가로 기존 구조물 하부를 통과하는 비개착 강관압입공법의 적용이 증가하는 추세이다. 비개착공법은 시공 중 기존구조물의 정상적인 운영 안정성이 확보되어야 하므로 강관압입에 의한 지표침하의 정밀한 예측이 필수적으로 필요한 공법이다. 강관압입 시 침하를 발생시키는 원인은 강관 선단과 강관과의 직경차에 의한 공극, 원활한 강관압입을 위한 과굴착, 강관과 지반과의 마찰에 의한 공극 발생 등이 있으며, 이는 Shield TBM 시공 시 발생하는 침하 원인과 유사하다. 본 연구에서는 Shield TBM의 침하 예측방법인 Gap Parameter Method와 Volume Loss Method를 이용하여 강관압입 시 침하를 예측하였으며, 현장시험을 통하여 예측방법에 대한 비교 분석을 수행하였다. 그 결과 Volume Loss 예측방법이 현장시험과 가장 유사한 결과로 나타났으나, 추후 예측방법의 Factor 결정 및 비개착공법 전체 침하예측을 위한 적용성 등 추가적인 연구가 필요할 것으로 판단된다. Non-excavation method is needed to secure the stability of existing structures during construction. Therefore, prediction of ground settlement is essential. Causes of settlement when using steel pipe indentation method are leading pipe-steel pipe gap, excessive excavation and soil-steel pipe friction etc. Also they are similar to the causes of settlement when using Shield TBM during construction. In this study, ground settlement during steel pipe indentation is predicted by the Gap Parameter Method and Volume Loss Method which are kinds of Shield TBM prediction Method. and compared with those of prediction methods by conducting field test. As a result, Volume Loss Prediction Method is the most similar to the field tests. However, It is needed to additional studies, such as decision of the factors and adaptability for total settlement predictions of non-excavation method.

The distribution patterns of shield tail gap and grouting pressure in soft ground is different from that in hard soil clays. In this paper, a two-dimensional nonlinear finite difference element model is proposed to study the influence of gap and grouting distribution to ground settlement. Based on shield tunneling project, a three-dimensional numerical model is proposed to research the mechanical behavior of tunnel construction, the supporting pressure of excavation face and the friction between shell and soil are taken into account during calculation. Results shows that: (1) Most of ground settlement is caused by the gap located at the shield tail, the distribution patterns of gap and the range of grouting have an important impact on the ground settlement in the cross ground section; (2) Disturbance to the stratum by support pressure and friction have a significant impact on the deformation in forms of bulge and settlement in the longitudinal ground section; (3) Based on the site-data, parameters reflecting the distribution of gap and grouting are obtained, the value and the range of ground settlement obtained by numerical calculation are very closed to that obtained by site monitoring ,which shows that this method can reveal and predict the ground settlement in soft clay during shield tunneling more realistically.

Diaphragm wall and internal support are commonly used forms of support for foundation pit engineering. The deformation of the foundation pit is generally considered during the excavation of the foundation pit, while the deformation in the stage of support removal is ignored. Based on the Shenzhen Vanke Coastal Foundation Pit Project, this paper mainly studies the impact of excavation and support removal of a foundation pit on the surrounding environment by finite element numerical simulation. The following conclusions are obtained. The deformation law of the support removal and the excavation is consistent, and it is obviously influenced by space-time effect. The amounts of deformation in the two stages are similar, and the sum of the deformation amounts accounts for more than 85% of the total deformation. The deformation of the bottom soil excavation and bottom support removal is the biggest. The reasons are as follows: firstly, the three supports are not equidistant, and the thickness of the bottom soil is large; Secondly, the thickness of granite residual soil varies greatly. Finally, after the support is removed, only the basement floor is provided with a change brace, which causes the cantilever section of the pile body to be long. The simulation results show that the lateral displacement at the top of the foundation pit, ground settlement and building settlement are basically consistent with the monitoring data. The results can provide reference for similar projects.

Accurate prediction of long-term ground settlement is essential for protecting existing structures and ensuring the safe progress of filling works. However, most existing methods fail to accurately predict the long-term deformation evolution following unexpected mutations. The paper proposes a mutation-aware hybrid model (MAHM) for forecasting long-term settlement with uncertainty quantified. The MAHM comprises two components: an ensemble empirical equation that supplies prior physical and engineering basis, and a modified long short-term memory network (LSTM) that refines the empirical model. The modified LSTM is constrained by a time-aware loss and employs mutation-aware bootstrap training to capture abrupt deformation features and to quantify predictive uncertainty. The method is validated in four challenging case studies and compared with three existing models. Results show that, on average, MAHM reduces the relative L 2 error by approximately 58.2%, 13.6%, and 22.1% relative to the ensemble empirical equation, LSTM, and an alternative multi-fidelity model, respectively; the mean prediction error of MAHM across scenarios remains within 20%, and the average coverage of actual values within the 95% confidence interval exceeds 80%. Additionally, MAHM's accuracy improves over time as it is dynamically updated. The model offers a practical and data-efficient approach for predicting long-term deformation, facilitating the timely implementation of mitigation measures.

Tunneling-induced settlement prediction using the hybrid feature selection method for feature optimization

Prediction of ground surface settlement by shield tunneling using XGBoost and Bayesian Optimization

Deformation behaviors and control indexes of metro-station deep excavations based on case histories

A thermal source model of cold metal transfer (CMT) welding is established, and the dynamic behavior of temperature field of CMT welding pool is simulated, and the corresponding distribution rule for temperature field is obtained. By combining the experimental measurement with ANSYS finite element numerical simulation, the thermal cycle curve of 16Mn steel surfacing welding is compared. The result of finite element calculation well consists with that of experimental measurement, demonstrating that this finite element numerical simulation model is reliable. The analysis of temperature field of CMT welding pool is reliably predicted, so as to offer a reference basis for simplifying the CMT welding experiment and choosing the parameters of CMT welding process.DOI: http://dx.doi.org/10.5755/j01.mech.22.4.16163

A numerical comparison of simplified Galerkin and machine learning reduced order models for vaginal deformations

Shield tunnels in operation are often affected by complex geological conditions, environmental factors, and structural aging, leading to cumulative damage in the segments and, consequently, increased deformation that compromises structural safety. To investigate the deformation behavior of tunnel linings under random damage conditions, this study integrates finite element numerical simulation with deep learning techniques to analyze and predict the deformation of shield tunnel segments. First, a refined three-dimensional finite element model was established, and a random damage modeling method was developed to simulate the deformation evolution of tunnel segments under different damage ratios. Additionally, a statistical analysis was conducted to assess the uncertainty in deformation caused by random damage. Furthermore, this study introduces a convolutional neural network (CNN) surrogate model to enable the rapid prediction of shield tunnel deformation under random damage conditions. The results indicate that as the damage ratio increases, both the mean deformation and its variability progressively rise, leading to increased deformation instability, demonstrating the cumulative effect of damage on segment deformation. Moreover, the 1D-CNN surrogate model was trained using finite element computation results, and predictions on the test dataset showed excellent agreement with FEM calculations. The surrogate model achieved a correlation coefficient (R2) exceeding 0.95 and an RMSE below 0.016 mm, confirming its ability to accurately predict the deformation of tunnel segments across different damage conditions. To the best of our knowledge, the finite-element–deep-learning hybrid approach proposed in this study provides a valuable theoretical foundation for predicting the deformation of in-service shield tunnels and assessing structural safety, offering scientific guidance for tunnel safety evaluation and damage repair strategies.

In this paper we study the accuracy of arbitrary subdomain reconstruction within finite element numerical simulations. We observe the influence of several parameters used to configure the boundary conditions at the reconstructed subdomain and how they shape the accuracy of the reconstructed values. We study several interpolation algorithms and different resolutions for the original and for the reconstructed simulation. We discuss our findings compared to real theoretical behavior and we summarize the circumstances, and parameters that have an impact on reconstructing arbitrary subdomains within finite element numerical simulations.