Urban underground space (UUS) development has been acknowledged as a positive contribution to urban resilience (UR). Such contribution has been qualitatively addressed in recent years, but only quantitatively discussed in few studies. Quantitative evaluation methods for UR are widely used in China and around the world, but the role of underground space is barely included. This paper provides a way to bridge this gap on the city scale. A UR evaluation framework was carefully constructed that covers the basic aspects and elements of UR. The contributions of UUS to UR were identified and integrated into the UR evaluation framework, and the measurement methods for each indicator related to UUS were determined. A case study of 19 sample cities in China were conducted using the integrated evaluation model. Correlation analysis and clustering analysis were further adopted to interpret the evaluation results, mainly with three indicators reflecting the level of UUS development, namely UUS area (m2), UUS density (104 m2/km2) and UUS area per capita (m2/person). The results showed a strong correlation between UUS area and UR. The average proportion of UR provided by UUS in the 19 sample cities was 16.46%, while the maximum figure reached 29.20%. The sample cities were clustered into four categories based on the relationship between the proportion of UR provided by UUS, UUS area, and GDP per capita, where both high and low UUS area tend to provide less proportion of resilience than the medium UUS area. Corresponding suggestions for UUS utilization were proposed to assist cities in achieving urban resilience.
The stability of underground entry-type excavations (UETEs) is of paramount importance for ensuring the safety of mining operations. As more engineering cases are accumulated, machine learning (ML) has demonstrated great potential for the stability evaluation of UETEs. In this study, a hybrid stacking ensemble method aggregating support vector machine (SVM), k-nearest neighbor (KNN), decision tree (DT), random forest (RF), multilayer perceptron neural network (MLPNN) and extreme gradient boosting (XGBoost) algorithms was proposed to assess the stability of UETEs. Firstly, a total of 399 historical cases with two indicators were collected from seven mines. Subsequently, to pursue better evaluation performance, the hyperparameters of base learners (SVM, KNN, DT, RF, MLPNN and XGBoost) and meta learner (MLPNN) were tuned by combining a five-fold cross validation (CV) and simulated annealing (SA) approach. Based on the optimal hyperparameters configuration, the stacking ensemble models were constructed using the training set (75% of the data). Finally, the performance of the proposed approach was evaluated by two global metrics (accuracy and Cohen’s Kappa) and three within-class metrics (macro average of the precision, recall and F1-score) on the test set (25% of the data). In addition, the evaluation results were compared with six base learners optimized by SA. The hybrid stacking ensemble algorithm achieved better comprehensive performance with the accuracy, Kappa coefficient, macro average of the precision, recall and F1-score were 0.92, 0.851, 0.885, 0.88 and 0.883, respectively. The rock mass rating (RMR) had the most important influence on evaluation results. Moreover, the critical span graph (CSG) was updated based on the proposed model, representing a significant improvement compared with the previous studies. This study can provide valuable guidance for stability analysis and risk management of UETEs. However, it is necessary to consider more indicators and collect more extensive and balanced dataset to validate the model in future.
Tunnelling has increasingly become an essential tool in the exploration of underground space. A typical construction problem is the face instability during tunnelling, posing a great threat to associated infrastructures. Tunnel face instability often occurs with the soil arching collapse. This study investigates the combined effect of cutterhead opening ratio and soil non-uniformity on soil arching effect and face stability, via conducting random finite-element analysis coupled with Monte-Carlo simulations. The results underscore that the face stability is strongly associated with the evolution of stress arch. The obtained stability factors in the uniform soils can serve as a reference for the design of support pressure in practical tunnelling engineering. In addition, non-uniform soils exhibit a lower stability factor than uniform soils, which implies that the latter likely yields an underestimated probability of face failure. The tunnel face is found to have a probability of failure more than 50% if the spatial non-uniformity of soil is ignored. In the end, a practical framework is established to determine factor of safety (FOS) corresponding to different levels of probability of face failure considering various opening ratios in non-uniform soils. The required FOS is 1.70 to limit the probability of face instability no more than 0.1%. Our findings can facilitate the prediction of probability of instability in the conventionally deterministic design of face pressure.
Contemporary demands necessitate the swift and accurate detection of cracks in critical infrastructures, including tunnels and pavements. This study proposed a transfer learning-based encoder-decoder method with visual explanations for infrastructure crack segmentation. Firstly, a vast dataset containing 7089 images was developed, comprising diverse conditions—simple and complex crack patterns as well as clean and rough backgrounds. Secondly, leveraging transfer learning, an encoder-decoder model with visual explanations was formulated, utilizing varied pre-trained convolutional neural network (CNN) as the encoder. Visual explanations were achieved through gradient-weighted class activation mapping (Grad-CAM) to interpret the CNN segmentation model. Thirdly, accuracy, complexity (computation and model), and memory usage assessed CNN feasibility in practical engineering. Model performance was gauged via prediction and visual explanation. The investigation encompassed hyperparameters, data augmentation, deep learning from scratch vs. transfer learning, segmentation model architectures, segmentation model encoders, and encoder pre-training strategies. Results underscored transfer learning's potency in enhancing CNN accuracy for crack segmentation, surpassing deep learning from scratch. Notably, encoder classification accuracy bore no significant correlation with CNN segmentation accuracy. Among all tested models, UNet-EfficientNet_B7 excelled in crack segmentation, harmonizing accuracy, complexity, memory usage, prediction, and visual explanation.
The emergence of curved shield tunnels poses a significant construction challenge. If the quality of the segment assembly is not guaranteed, many segment cracks and damage will result from the stress concentration. Sensing the contact stresses between segmental joints is necessary to improve the quality of segments assembled for shield tunnel construction. Polyvinylidene difluoride (PVDF) piezoelectric material was chosen for the sensor because it can convert contact stresses into electrical signals, allowing the state of the segmental joints to be effectively sensed. It matches the working environment between the segmental joints of the shield tunnel, where flexible structures such as rubber gaskets and force transfer pads are present. This study proposes a piezoelectric sensing method for segmental joints in shield tunnels and conducts laboratory tests, numerical analyses, and field tests to validate the feasibility of the method. The results indicate that the PVDF film sensor can effectively sense the entire compression process of the gasket with different amounts of compression. The piezoelectric cable sensor can effectively sense the joint offset direction of the gasket. For differently shaped sections, the variation in the force sensed by the piezoelectric cable sensors was different, as verified by numerical simulation. Through the field test, it was found that the average contact stress between the segmental joints was in the range of 1.2-1.8 MPa during construction of the curved shield tunnels. The location of the segmental joints and the type of segment affect the contact stress value. The field monitoring results show that piezoelectric sensing technology can be successfully applied during assembly of the segments for effective sensing of the contact stress.
Ground penetrating radar (GPR) is a vital non-destructive testing (NDT) technology that can be employed for detecting the backfill grouting of shield tunnels. To achieve intelligent analysis of GPR data and overcome the subjectivity of traditional data processing methods, the CatBoost & BO-TPE model was constructed for regressing the grouting thickness based on GPR waveforms. A full-scale model test and corresponding numerical simulations were carried out to collect GPR data at 400 and 900 MHz, with known backfill grouting thickness. The model test helps address the limitation of not knowing the grout body condition in actual field detection. The data were then used to create machine learning datasets. The method of feature selection was proposed based on the analysis of feature importance and the electromagnetic (EM) propagation law in mediums. The research shows that: (1) the CatBoost & BO-TPE model exhibited outstanding performance in both experimental and numerical data, achieving R2 values of 0.9760, 0.8971, 0.8808, and 0.5437 for numerical data and test data at 400 and 900 MHz. It outperformed extreme gradient boosting (XGBoost) and random forest (RF) in terms of performance in the backfill grouting thickness regression; (2) compared with the full-waveform GPR data, the feature selection method proposed in this paper can promote the performance of the model. The selected features within the 5-30 ns of the A-scan can yield the best performance for the model; (3) compared to GPR data at 900 MHz, GPR data at 400 MHz exhibited better performance in the CatBoost & BO-TPE model. This indicates that the results of the machine learning model can provide feedback for the selection of GPR parameters; (4) the application results of the trained CatBoost & BO-TPE model in engineering are in line with the patterns observed through traditional processing methods, yet they demonstrate a more quantitative and objective nature compared to the traditional method.
Geotechnical centrifuge tests were conducted to examine the influence of invert voids and surface traffic loads on 1400 mm diameter reinforced concrete pipes buried with a shallow soil cover depth of 700 mm. Void formation beneath the pipe was simulated during centrifuge testing. The test results revealed that before void formation, the surface load directly above the middle of the pipe caused a significant increase in not only the circumferential bending moments but also the longitudinal bending moments, the latter of which was considerable and could not be ignored. Void formation beneath the middle of the pipe led to a reduction in both the circumferential bending moments and longitudinal bending moments at all measuring positions, i.e., crown, springline, and invert. The most significant reduction occurred at the invert, and there was even a reversal in the sign of the invert longitudinal bending moment. A comparison was made between centrifuge tests with erosion voids and surface loads at different horizontal positions, which had a marked influence even when the positions differed by half a pipe length. Joint rotation played an important role in relieving large bending moments of pipe barrels in a jointed pipeline when the void and surface load were located at the joint.
Over the last decades, an expansion of the underground network has been taking place to cope with the increasing amount of moving people and freight. As a consequence, it is of vital importance to guarantee the full functionality of the tunnel network by means of preventive maintenance and the monitoring of the tunnel lining state over time. A new method has been developed for the real-time prediction of the utilization level in tunnel segmental linings based on input monitoring data. The new concept is founded on a framework, which encompasses an offline and an online stage. In the former, the generation of feedforward neural networks is accomplished by employing synthetically produced data. Finite element simulations of the lining structure are conducted to analyze the structural response under multiple loading conditions. The scenarios are generated by assuming ranges of variation of the model input parameters to account for the uncertainty due to the not fully determined in situ conditions. Input and target quantities are identified to better assess the structural utilization of the lining. The latter phase consists in the application of the methodological framework on input monitored data, which allows for a real-time prediction of the physical quantities deployed for the estimation of the lining utilization. The approach is validated on a full-scale test of segmental lining, where the predicted quantities are compared with the actual measurements. Finally, it is investigated the influence of artificial noise added to the training data on the overall prediction performances and the benefits along with the limits of the concept are set out.
Calculating the parameters of the ground shock induced by an underground explosion is a complex energy coupling problem. It is difficult to establish a unified ground shock coupling law from limited test data. This paper summarizes the research results obtained at home and abroad and systematically analyzes the coupling mechanism of craters formed by an underground explosion and the ground shock. The differences between the concepts of “closed-explosion critical depth” and “equivalent closed-explosion critical depth” are clearly explained. The spreading of the ground shock energy is attributed to the explosive expansion of the air cavity, revealing a linear relationship between the volume of the cavity region (or the volume of the crack region) and the ground shock energy associated with the underground explosion. The proportionality factor is related to the mechanical properties of the medium and is independent of the magnitude of the explosion equivalent. Based on this, a theoretical calculation formula and conversion method for the ground shock coupling coefficient were established. Explosion tests were conducted in clay and Plexiglass under varying burial depths. The test results were consistent with the theoretically calculated results. Our study provides a theoretical basis for the design of explosion-resistant structures in underground engineering.
Ground losses due to tunneling would induce settlement of nearby raft foundations. To study the change in behavior of the raft foundations over time due to tunnel excavation in soft clay, a series of centrifuge model tests were conducted. The results reveal that the raft stiffness has a significant influence on the development of the gap between the raft and the ground. The width of the gap beneath the flexible foundation would increase over time, leading to a further increase in tensile strain after excavation, whereas the gap for raft foundations with a large stiffness would reduce with time, causing a gradual decrease in tensile strain. The modification factor (MF) design approach is also evaluated with the test results and demonstrates that the MF design approach would underestimate the tensile strain of the flexible raft and provide a relatively conservative prediction for larger stiffnesses.
Parameters of foam penetration in earth pressure balance (EPB) shield tunnelling, such as permeability coefficients and penetration distances, significantly impact tunnel face stability. However, existing studies have faced inaccuracies in analysing these parameters due to imitations in experimental methods. This study addresses this issue by employing enhanced methods for a more precise analysis of foam penetration. Experiments involving three distinct sand types (coarse, medium, and fine) and three foam expansion ratios (FER) (10, 15, and 20) are conducted using a modified model test setup. Benefiting from a novel computer vision-based method, the model test outcomes unveil two distinct foam penetration paths: liquid migration (Lw) and bubble migration (Lf). Three penetration phases — namely, injection, blockage & drainage, and breakage — are identified based on Lw and Lf variations. The initial “injection” phase conforms to Darcy's law and is amenable to mathematical description. The foam with FER of 15 has the maximum viscosity and, hence the Lf and permeability in the penetration tests with FER of 15 are the lowest in the same sand. The bubble size distribution of foam with different FER shows minor differences. Nevertheless, the characteristics of foam penetration vary due to the distinct particle size distribution (PSD) of different sands. Foam penetration creates low-permeability layers in both medium and fine sands due to the larger bubble size of the foam compared to the estimated pore sizes of medium and fine sands. While the coarse sand results in a different situation due to its large pore size. The distinctive characteristics of foam penetration in different sand strata are notably shaped by FER, PSD, and pore size distributions. These insights shed light on the complex interactions during foam penetration at the tunnel face, contributing valuable knowledge to EPB shield tunnelling practices.
Analytics and visualization of multi-dimensional and complex geo-data, such as three-dimensional (3D) subsurface ground models, is critical for development of underground space and design and construction of underground structures (e.g., tunnels, dams, and slopes) in engineering practices. Although complicated 3D subsurface ground models now can be developed from site investigation data (e.g., boreholes) which is often sparse in practice, it remains a great challenge to visualize a 3D subsurface ground model with sophisticated stratigraphic variations by conventional two-dimensional (2D) geological cross-sections. Virtual reality (VR) technology, which has an attractive capability of constructing a virtual environment that links to the physical world, has been rapidly developed and applied to visualization in various disciplines recently. Leveraging on the rapid development of VR, this study proposes a framework for immersive visualization of 3D subsurface ground models in geo-applications using VR technology. The 3D subsurface model is first developed from limited borehole data in a data-driven manner. Then, a VR system is developed using related software and hardware devices currently available in the markets for immersive visualization and interaction with the developed 3D subsurface ground model. The results demonstrate that VR visualization of the 3D subsurface ground model in an immersive environment has great potential in revolutionizing the geo-practices from 2D cross-sections to a 3D immersive virtual environment in digital era, particularly for the emerging digital twins.
Subjected to the coupling action of multiple hazards in hydraulic engineering, hydraulic tunnels may be corroded and damaged to varying degrees during their service lives, which will decrease the seismic performance of these structures. However, the research and seismic design of significant hydraulic engineering projects focus on investigating the structural response based on the design material parameters, which may overestimate the seismic capacity of structures during their service lives. In this paper, research is performed to identify the effect of hydro-chemo-mechanical corrosion on the seismic performance of hydraulic tunnels with different burial depths. A plastic damage model of time-varying concrete degradation induced by the hydro-chemo-mechanical effect is first determined and implemented, and the endurance time acceleration records are generated in MATLAB. Then, a study of the endurance time relationship of hydro-chemo-mechanical corrosion-affected hydraulic tunnels, considering the fluid-structure-surrounding rock interaction systems throughout the service period, is undertaken to directly associate the structural response with the predefined evaluation index. Moreover, this research constructs 3D time-varying fragility surfaces considering the hydro-chemo-mechanical effect and seismic intensity. The results show that the relative displacement of hydro-chemo-mechanical corrosion-affected hydraulic tunnels is larger than that of nonaffected hydraulic tunnels. Hydro-chemo-mechanical effect-induced material deterioration will lead to an increase in the cumulative damage (crack) area and damage degree of hydraulic tunnels. Additionally, the seismic fragility analysis shows that the longer the service time of hydro-chemo-mechanical corrosion-affected hydraulic tunnels, the more likely they are to collapse. Hence, attention should be given to improving the aseismic capacity of hydro-chemo-mechanical corrosion-affected hydraulic tunnels in future seismic design and performance assessments.
We conducted a study to evaluate the potential and robustness of gradient boosting algorithms in rock burst assessment, established a variational autoencoder (VAE) to address the imbalance rock burst dataset, and proposed a multilevel explainable artificial intelligence (XAI) tailored for tree-based ensemble learning. We collected 537 data from real-world rock burst records and selected four critical features contributing to rock burst occurrences. Initially, we employed data visualization to gain insight into the data's structure and performed correlation analysis to explore the data distribution and feature relationships. Then, we set up a VAE model to generate samples for the minority class due to the imbalanced class distribution. In conjunction with the VAE, we compared and evaluated six state-of-the-art ensemble models, including gradient boosting algorithms and the classical logistic regression model, for rock burst prediction. The results indicated that gradient boosting algorithms outperformed the classical single models, and the VAE-classifier outperformed the original classifier, with the VAE-NGBoost model yielding the most favorable results. Compared to other resampling methods combined with NGBoost for imbalanced datasets, such as synthetic minority oversampling technique (SMOTE), SMOTE-edited nearest neighbours (SMOTE-ENN), and SMOTE-tomek links (SMOTE-Tomek), the VAE-NGBoost model yielded the best performance. Finally, we developed a multilevel XAI model using feature sensitivity analysis, Tree Shapley Additive exPlanations (Tree SHAP), and Anchor to provide an in-depth exploration of the decision-making mechanics of VAE-NGBoost, further enhancing the accountability of tree-based ensemble models in predicting rock burst occurrences.
The confined aquifer dewatering for long-deep excavations usually encounters challenges due to complicated geotechnical conditions, large excavation sizes, and high hydraulic pressures. To propose the most efficient scheme of confined aquifer dewatering for long-deep excavations, dewatering optimizations were performed using the simulation-optimization method. An open cut tunnel of the Jiangyin-Jingjiang Yangtze River Tunnel Project was taken as an example. The methods of finite element and linear programming (LP) were combined to optimize the dewatering process. A three-dimensional finite element model was developed. After simulating the pumping tests, hydraulic conductivity was inverted. Then, necessary parameters in the LP method were determined by simulating dewatering with each pumping well, and various LP models were developed based on some important influence factors such as dewatering sequence, considered pumping wells, and pumping rate limitation. Finally, the optimal pumping rates were solved and applied to the numerical model, with induced drawdown and ground settlement computed for comparison. The results indicate that the optimization can significantly reduce the required wells in the original design. Dewatering in the deepest zone exhibits the highest efficiency for long-deep excavations with gradually varying depths. For the dewatering sequence from the shallowest to the deepest zone, more pumping wells are required but less energy is consumed. Higher quantity and more advantageous locations of pumping wells in the LP model usually result in lower total pumping rate, drawdown, and ground settlement. If more pumping wells are considered in the deepest zone, pumping rate limitation of single well will only slightly increase the total pumping rate, number of required pumping wells, drawdown, and ground settlement.
Ovaling deformation of circular tunnels has received great interest from the tunneling community because this mode of seismic-induced deformation is considered the most critical. However, there is growing evidence that other deformation modes can also be important and thus need to be considered in design. This study presents a new analytical solution to estimate axial bending (snaking), a mode of deformation caused by S-waves impinging on a tunnel parallel to the tunnel axis. The solution is developed using the soil-structure interaction approach with the assumption that the interface between the ground and the tunnel lining is frictionless (full-slip). Full dynamic numerical simulations are conducted to verify the new full-slip solution, together with the existing no-slip solution. Effects of dynamic amplification are also explored for both full-slip and no-slip interface conditions by changing the wavelength (or frequency) of the seismic input motions.
In the longitudinal seismic deformation method for shield tunnels, one of the most commonly used is the longitudinal equivalent stiffness beam model (LES) for simulating the mechanical behavior of the lining. In this model, axial deformation and bending deformation are independent, so the equivalent stiffness is a constant value. However, the actual situation is that axial deformation and bending deformation occur simultaneously, which is not considered in LES. At present, we are not clear about the effect on the calculation results when axial deformation and bending deformation occur simultaneously. Therefore, in this paper, we improve the traditional LES by taking the relative deformation as a load and considering the coordinated deformation of axial and bending degrees of freedom. This improved model is called DNLES, and its neutral axis equations are an explicit expression. Then, we propose an iterative algorithm to solve the calculation model of the DNLES-based longitudinal seismic deformation method. Through a calculation example, we find that the internal forces based on LES are notably underestimated than those of DNLES in the compression bending zone, while are overestimated in the tension bending zone. When considering the combined effect, the maximum bending moment reached 13.7 times that of the LES model, and the axial pressure and tension were about 1.14 and 0.96 times, respectively. Further analysis reveals the coordinated deformation process in the axial and bending directions of the shield tunnel, which leads to a consequent change in equivalent stiffness. This explains why, in the longitudinal seismic deformation method, the traditional LES may result in unreasonable calculation results.
Discontinuity is critical for strength, deformability, and permeability of rock mass. Set information is one of the essential discontinuity characteristics and is usually accessed by orientation grouping. Traditional methods of identifying optimal discontinuity set numbers are usually achieved by clustering validity indexes, which mainly relies on the aggregation and dispersion of clusters and leads to the inaccuracy and instability of evaluation. This paper proposes a new method of Fisher mixed evaluation (FME) to identify optimal group numbers of rock mass discontinuity orientation. In FME, orientation distribution is regarded as the superposition of Fisher mixed distributions. Optimal grouping results are identified by considering the fitting accuracy of Fisher mixed distributions, the probability monopoly and central location significance of independent Fisher centers. A Halley-Expectation-Maximization (EM) algorithm is derived to achieve an automatic fitting of Fisher mixed distribution. Three real rock discontinuity models combined with three orientation clustering algorithms are adopted for discontinuity grouping. Four clustering validity indexes are used to automatically identify optimal group numbers for comparison. The results show that FME is more accurate and robust than the other clustering validity indexes in optimal discontinuity group number identification for different rock models and orientation clustering algorithms.
A novel coupled model integrating Elman-AdaBoost with adaptive mutation sparrow search algorithm (AM-SSA), called AMSSA-Elman-AdaBoost, is proposed for predicting the existing metro tunnel deformation induced by adjacent deep excavations in soft ground. The novelty is that the modified SSA proposes adaptive adjustment strategy to create a balance between the capacity of exploitation and exploration. In AM-SSA, firstly, the population is initialized by cat mapping chaotic sequences to improve the ergodicity and randomness of the individual sparrow, enhancing the global search ability. Then the individuals are adjusted by Tent chaotic disturbance and Cauchy mutation to avoid the population being too concentrated or scattered, expanding the local search ability. Finally, the adaptive producer-scrounger number adjustment formula is introduced to balance the ability to seek the global and local optimal. In addition, it leads to the improved algorithm achieving a better accuracy level and convergence speed compared with the original SSA. To demonstrate the effectiveness and reliability of AM-SSA, 23 classical benchmark functions and 25 IEEE Congress on Evolutionary Computation benchmark test functions (CEC2005), are employed as the numerical examples and investigated in comparison with some well-known optimization algorithms. The statistical results indicate the promising performance of AM-SSA in a variety of optimization with constrained and unknown search spaces. By utilizing the AdaBoost algorithm, multiple sets of weak AMSSA-Elman predictor functions are restructured into one strong predictor by successive iterations for the tunnel deformation prediction output. Additionally, the on-site monitoring data acquired from a deep excavation project in Ningbo, China, were selected as the training and testing sample. Meanwhile, the predictive outcomes are compared with those of other different optimization and machine learning techniques. In the end, the obtained results in this real-world geotechnical engineering field reveal the feasibility of the proposed hybrid algorithm model, illustrating its power and superiority in terms of computational efficiency, accuracy, stability, and robustness. More critically, by observing data in real time on daily basis, the structural safety associated with metro tunnels could be supervised, which enables decision-makers to take concrete control and protection measures.