An evolutionary adaptive neuro-fuzzy inference system for estimating field penetration index of tunnel boring machine in rock mass

Field penetration index (FPI) is one of the representative key parameters to examine the tunnel boring machine (TBM) performance. Lack of accurate FPI prediction can be responsible for numerous disastrous incidents associated with rock mechanics and engineering. This study aims to predict TBM performance (i.e. FPI) by an efficient and improved adaptive neuro-fuzzy inference system (ANFIS) model. This was done using an evolutionary algorithm, i.e. artificial bee colony (ABC) algorithm mixed with the ANFIS model. The role of ABC algorithm in this system is to find the optimum membership functions (MFs) of ANFIS model to achieve a higher degree of accuracy. The procedure and modeling were conducted on a tunnelling database comprising of more than 150 data samples where brittleness index (BI), fracture spacing, α angle between the plane of weakness and the TBM driven direction, and field single cutter load were assigned as model inputs to approximate FPI values. According to the results obtained by performance indices, the proposed ANFIS_ABC model was able to receive the highest accuracy level in predicting FPI values compared with ANFIS model. In terms of coefficient of determination (R²), the values of 0.951 and 0.901 were obtained for training and testing stages of the proposed ANFIS_ABC model, respectively, which confirm its power and capability in solving TBM performance problem. The proposed model can be used in the other areas of rock mechanics and underground space technologies with similar conditions.     

Real-time prediction of rock mass classification based on TBM operation big data and stacking technique of ensemble learning

Real-time prediction of the rock mass class in front of the tunnel face is essential for the adaptive adjustment of tunnel boring machines(TBMs).During the TBM tunnelling process,a large number of operation data are generated,reflecting the interaction between the TBM system and surrounding rock,and these data can be used to evaluate the rock mass quality.This study proposed a stacking ensemble classifier for the real-time prediction of the rock mass classification using TBM operation data.Based on the Songhua River water conveyance project,a total of 7538 TBM tunnelling cycles and the corresponding rock mass classes are obtained after data preprocessing.Then,through the tree-based feature selection method,10 key TBM operation parameters are selected,and the mean values of the 10 selected features in the stable phase after removing outliers are calculated as the inputs of classifiers.The preprocessed data are randomly divided into the training set(90%)and test set(10%)using simple random sampling.Besides stacking ensemble classifier,seven individual classifiers are established as the comparison.These classifiers include support vector machine(SVM),k-nearest neighbors(KNN),random forest(RF),gradient boosting decision tree(GBDT),decision tree(DT),logistic regression(LR)and multilayer perceptron(MLP),where the hyper-parameters of each classifier are optimised using the grid search method.The prediction results show that the stacking ensemble classifier has a better performance than individual classifiers,and it shows a more powerful learning and generalisation ability for small and imbalanced samples.Additionally,a relative balance training set is obtained by the synthetic minority oversampling technique(SMOTE),and the influence of sample imbalance on the prediction performance is discussed.     

Multiclass stand-alone and ensemble machine learning algorithms utilised to classify soils based on their physico-chemical characteristics

This study has provided an approach to classify soil using machine learning. Multiclass elements of stand-alone machine learning algorithms (i.e. logistic regression (LR) and artificial neural network (ANN)), decision tree ensembles (i.e. decision forest (DF) and decision jungle (DJ)), and meta-ensemble models (i.e. stacking ensemble (SE) and voting ensemble (VE)) were used to classify soils based on their intrinsic physico-chemical properties. Also, the multiclass prediction was carried out across multiple cross-validation (CV) methods, i.e. train validation split (TVS), k-fold cross-validation (KFCV), and Monte Carlo cross-validation (MCCV). Results indicated that the soils' clay fraction (CF) had the most influence on the multiclass prediction of natural soils' plasticity while specific surface and carbonate content (CC) possessed the least within the nature of the dataset used in this study. Stand-alone machine learning models (LR and ANN) produced relatively less accurate predictive performance (accuracy of 0.45, average precision of 0.5, and average recall of 0.44) compared to tree-based models (accuracy of 0.68, average precision of 0.71, and recall rate of 0.68), while the meta-ensembles (SE and VE) outperformed (accuracy of 0.75, average precision of 0.74, and average recall rate of 0.72) all the models utilised for multiclass classification. Sensitivity analysis of the meta-ensembles proved their capacities to discriminate between soil classes across the methods of CV considered. Machine learning training and validation using MCCV and KFCV methods enabled better prediction while also ensuring that the dataset was not overfitted by the machine learning models. Further confirmation of this phenomenon was depicted by the continuous rise of the cumulative lift curve (LC) of the best performing models when using the MCCV technique. Overall, this study demonstrated that soil's physico-chemical properties do have a direct influence on plastic behaviour and, therefore, can be relied upon to classify soils.     

Improved prediction of slope stability using a hybrid stacking ensemble method based on finite element analysis and field data

Slope failures lead to catastrophic consequences in numerous countries and thus the stability assessment for slopes is of high interest in geotechnical and geological engineering researches.A hybrid stacking ensemble approach is proposed in this study for enhancing the prediction of slope stability.In the hybrid stacking ensemble approach,we used an artificial bee colony(ABC)algorithm to find out the best combination of base classifiers(level 0)and determined a suitable meta-classifier(level 1)from a pool of 11 individual optimized machine learning(OML)algorithms.Finite element analysis(FEA)was conducted in order to form the synthetic database for the training stage(150 cases)of the proposed model while 107 real field slope cases were used for the testing stage.The results by the hybrid stacking ensemble approach were then compared with that obtained by the 11 individual OML methods using confusion matrix,F1-score,and area under the curve,i.e.AUC-score.The comparisons showed that a significant improvement in the prediction ability of slope stability has been achieved by the hybrid stacking ensemble(AUC?90.4%),which is 7%higher than the best of the 11 individual OML methods(AUC?82.9%).Then,a further comparison was undertaken between the hybrid stacking ensemble method and basic ensemble classifier on slope stability prediction.The results showed a prominent performance of the hybrid stacking ensemble method over the basic ensemble method.Finally,the importance of the variables for slope stability was studied using linear vector quantization(LVQ)method.     

Utilizing rock mass properties for predicting TBM performance in hard rock condition

The key parameters on the estimation of tunnel-boring machine (TBM) performance are rock strength, toughness, discontinuity in rock mass, type of TBM and its specifications. The aim of this study is to both assess the influence of rock mass properties on TBM performance and construct a new empirical equation for estimation of the TBM performance. To achieve this aim, the database composed of actual measured TBM penetration rate and rock properties (i.e., uniaxial compressive strength, Brazilian tensile strength, rock brittleness/toughness, distance between planes of weakness, and orientation of discontinuities in rock mass) were established using the data collected from one hard rock TBM tunnel (the Queens Water Tunnel # 3, Stage 2) about 7.5 km long, New York City, USA. Intact rock properties were obtained from laboratory studies conducted at the Earth Mechanics Institute (EMI) in the Colorado School of Mines, CO, USA. Based on generated database, the statistical analyses were performed between available rock properties and measured TBM data in the field. The result revealed that rock mass properties have strong affect on TBM performance. It is concluded that TBM performance could be estimated as a function of rock properties utilizing new equation (r = 0.82).     

New developments in TBM tunnelling for changing grounds

Mechanized tunnelling has seen substantial growth in the recent years in the construction industry for small to large scale infrastructure works. This development is due to a growing number of large-scale projects that were successfully realized with certainty and reliability in construction, compliance with construction schedules and thus supported the economic success of the projects. Tunnel projects are planned and implemented today in changing grounds that could not be realized without the newest technological progress and innovations in mechanized tunnelling. Today’s demand in tunnelling is to construct tunnels with high safety standards especially in urban areas with sensitive geological and hydrogeological conditions. In the context of this publication, new developments in TBM tunnelling related to changing ground conditions are highlighted. References of major infrastructure projects will be addressed that illustrate also the effect and importance of mechanized tunnelling technology used today to complete infrastructure projects on time and with high quality and technical standards.     

Probing ahead of TBMs by geophysical means

The potential of geophysical probing methods in TBM tunnelling is discussed. Modern TBMs have made it possible to tunnel through a wide range of geological conditions. However, the development towards more complicated machines has raised prices and often causes delays before the machines can begin operating properly. If a reliable system were available, the machine could be given a simpler design and the tunnelling process could be continuously adjusted to the prevailing ground conditions without jeopardizing safety. Geophysical investigation methods such as seismics and radar now offer the possibility of monitoring ground conditions ahead of the tunnel face. An analysis of a number of TBM projects has shown that if certain requirements on the range and time of the investigations are fulfilled, probing can be integrated into the tunnelling cycle.     

Tunnelling through a frequently changing and mixed ground: A case history in Singapore

The Kranji tunnel is part of the Deep Tunnel Sewerage System in Singapore. It is approximately 12.6 km in length. Along the tunnel alignment, all the ground is composed of granite with different weathering grades (from fresh rock to residual soil). The changing ground from hard rock to mixed face and soft ground (and vice versa) at the tunnel level was anticipated. The tunnel depth along the route is between 15 m and 50 m. Two EPB TBMs were deployed at this tunnel with a bored diameter 4.90 m. These machines were designed so that both hard rock and soft ground could be excavated. The cutter head was equipped with a combination of both rippers and disc cutters. During the excavation, it was found that the frequency of the ground change between hard rock and residual soil is much higher than that expected. Due to the frequently changing ground, correspondingly the tunnel boring machine (TBM) operation mode had to be transferred frequently from hard rock tunnelling to transition mode and to earth pressure balance (EPB) close mode. It resulted in great difficulties for the TBM in an optimized operation condition. These difficulties included high cutter wear and flat cutters, tunnel face instability, water inflow at weathering interface, and time delays. In order to overcome these problems and speed up the tunnelling progress, the TBM used in the north drive was modified to attempt the frequently changing ground. The performance of the modified TBM was highly improved. However, the highly abrasive and frequently changing mixed face ground still caused high cutter wear, especially flat cutter wear. These posed many challenges to the equipment and the tunnel crew.     

Prediction of flyrock distance induced by mine blasting using a novel Harris Hawks optimization-based multi-layer perceptron neural network

In mining or construction projects, for exploitation of hard rock with high strength properties, blasting is frequently applied to breaking or moving them using high explosive energy. However, use of explosives may lead to the flyrock phenomenon. Flyrock can damage structures or nearby equipment in the surrounding areas and inflict harm to humans, especially workers in the working sites. Thus, prediction of flyrock is of high importance. In this investigation, examination and estimation/forecast of flyrock distance induced by blasting through the application of five artificial intelligent algorithms were carried out. One hundred and fifty-two blasting events in three open-pit granite mines in Johor, Malaysia, were monitored to collect field data. The collected data include blasting parameters and rock mass properties. Site-specific weathering index (WI), geological strength index (GSI) and rock quality designation (RQD) are rock mass properties. Multi-layer perceptron (MLP), random forest (RF), support vector machine (SVM), and hybrid models including Harris Hawks optimization-based MLP (known as HHO-MLP) and whale optimization algorithm-based MLP (known as WOA-MLP) were developed. The performance of various models was assessed through various performance indices, including a10-index, coefficient of determination (R²), root mean squared error (RMSE), mean absolute percentage error (MAPE), variance accounted for (VAF), and root squared error (RSE). The a10-index values for MLP, RF, SVM, HHO-MLP and WOA-MLP are 0.953, 0.933, 0.937, 0.991 and 0.972, respectively. R² of HHO-MLP is 0.998, which achieved the best performance among all five machine learning (ML) models.     

On the use of 1g physical models for ground movements and soil-structure interaction problems

The paper focusses on the use of physical modelling in ground movements (induced by underground cavity collapse or mining/tunnelling) and associated soil-structure interaction issues. The paper presents first an overview of using 1g physical models to solve geotechnical problems and soil-structure interactions related to vertical ground movements. Then the 1g physical modelling application is illustrated to study the development of damage in masonry structure due to subsidence and cavity collapse. A large-scale 1g physical model with a 6 m³ container and 15 electric jacks is presented with the use of a three-dimensional (3D) image correlation technique. The influence of structure position on the subsidence trough is analysed in terms of crack density and damage level. The obtained results can improve the methodology and practice for evaluation of damage in masonry structures. Nevertheless, ideal physical model is difficult to achieve. Thus, future improvement of physical models (analogue materials and instrumentation) could provide new opportunities for using 1g physical models in geotechnical and soil-structure applications and research projects.     

Correlation of tunnel convergence with TBM operational parameters and chip size in the Ghomroud tunnel, Iran

Evaluating the impact of rock mass properties on a tunneling operation is crucial, especially when using a tunnel boring machine (TBM). It is an integral part of machine selection and performance prediction in the design and bidding stage. Monitoring and analysis of ground conditions during the construction is also essential to allow the operator to take precautionary measures in adverse geological conditions. This involves adjusting TBM operational parameters such as machine thrust and penetration to avoid potential problems caused by face collapse or excessive convergence and subsequent machine seizure that can cause long delays. Tunnel wall convergence is a function of rock mass characteristics, in situ stresses, size of excavation, and rate of penetration (ROP). It is one of the main factors in determining the use of shielded machines in deep rock tunnel projects. The case study of the Ghomroud water conveyance tunnel project, under construction by a double shield TBM, is used to examine the effect of rock mass parameters on tunnel convergence and hence on the need for over excavation and shield lubrication to avoid problems such as shield seizure. Results of a preliminary analysis of field observations show that the amount of the tunnel convergence can have a direct relationship with the percentage of powder and large rock fragments in the muck. In addition, tunnel convergence has shown a strong relationship with the TBM thrust/torque and rate of penetration (ROP). These relationships have been examined and the results of the analysis as well as the resulting formulas will be explained in this paper.     

Theoretical analysis and seismic investigation for TBM jamming in squeezing fissile slate

The S tunnel is a 4.2 km-long headrace tunnel. In the tunnelling project, the ground was assumed to be hard slate and suitable for TBM excavation based on the primary site investigation. However, TBM jamming frequently occurred with the increase of the tunnel cover, and the TBM excavation was cancelled. In order to investigate the TBM jamming, theoretical analyses and seismic investigations were conducted. It was found that analytical model proposed in this paper well explained the influence of the cover on the possibility of TBM jamming. It was also found that the depth of the loosened zone was expanded 6–8 m at the location where TBM jamming occurred.     

Performance prediction of hard rock Tunnel Boring Machines (TBMs) in difficult ground

Performance prediction of TBMs is an essential part of project scheduling and cost estimation. This process involves a good understanding of the complexities in the site geology, machine specification, and site management. Various approaches have been used over the years to estimate TBM performance in a given ground condition, many of them were successful and within an acceptable range, while some missing the actual machine performance by a notable margin. Experience shows that the best approach for TBM performance prediction is to use various models to examine the range of estimated machine penetration and advance rates and choose a rate that best represents the working conditions that is closest to the setting of the model used for the estimation. This allows the engineers to avoid surprises and to identify the parameters that could dominate machine performance in each case. This paper reviews the existing models for performance prediction of TBMs and some of the ongoing research on developing better models for improved accuracy of performance estimate and increasing TBM utilization.     

Tunnel boring machines under squeezing conditions

Squeezing ground represents a challenging operating environment as it may slow down or obstruct TBM operation. Due to the geometrical constraints of the equipment, relatively small convergences of one or two decimetres may lead to considerable difficulties in the machine area (sticking of the cutter head, jamming of the shield) or in the back-up area (e.g., jamming of the back-up equipment, inadmissible convergences of the bored profile, damage to the tunnel support). Depending on the number and the length of the critical stretches, squeezing conditions may even call into question the feasibility of a TBM drive. This paper sets out firstly to give an overview of the specific problems of TBM tunnelling under squeezing conditions; secondly to analyse the factors governing TBM performance by means of a structured examination of the multiple interfaces and interactions between ground, tunnelling equipment and support; and thirdly to provide a critical review of the technical options existing or proposed for coping with squeezing ground in mechanized tunnelling.     

A quantitative evaluation of the sustainability or unsustainability of three tunnelling projects

There is a perception that tunnelling is sustainable. This is because it occurs underground, and consequently does not significantly interfere with surface or atmospheric processes unlike other anthropogenic activities. However, the tools and assessments used in tunnelling projects to evaluate sustainability in the construction and operational phases are primarily concerned with the reduction of carbon footprint and environmental performance. This does not provide a suitable approach to determining the sustainability of a tunnelling project directly. Environmental Impact Assessment (EIA) on the other hand does have this potential. However, it requires two things: (1) a suitable quantitative-based method of EIA; and most critically and (2) a means to evaluate sustainability from the EIA results. Based upon the recent work of Namin et al. (2014) concerning a new EIA methodology for tunnelling projects, this paper applies an established mathematical model of sustainability to the results of the EIA to determine the sustainability or unsustainability of tunnelling projects. The model’s application, in the form of an algorithm, evaluates three case studies assessed by Namin et al. (2014). The results are analysed and discussed in respect to the three projects’ construction and operation phases. The broader context of the results is then discussed in respect to the use of underground space as a means to achieve sustainability.     

Modeling approaches and performance for estimating personal exposure to household air pollution: A case study in Kenya

This study assessed the performance of modeling approaches to estimate personal exposure in Kenyan homes where cooking fuel combustion contributes substantially to household air pollution (HAP). We measured emissions (PM_2.5, black carbon, CO); household air pollution (PM_2.5, CO); personal exposure (PM_2.5, CO); stove use; and behavioral, socioeconomic, and household environmental characteristics (eg, ventilation and kitchen volume). We then applied various modeling approaches: a single-zone model; indirect exposure models, which combine person-location and area-level measurements; and predictive statistical models, including standard linear regression and ensemble machine learning approaches based on a set of predictors such as fuel type, room volume, and others. The single-zone model was reasonably well-correlated with measured kitchen concentrations of PM_2.5 (R² = 0.45) and CO (R² = 0.45), but lacked precision. The best performing regression model used a combination of survey-based data and physical measurements (R² = 0.76) and a root mean-squared error of 85 µg/m³, and the survey-only-based regression model was able to predict PM2.5 exposures with an R² of 0.51. Of the machine learning algorithms evaluated, extreme gradient boosting performed best, with an R² of 0.57 and RMSE of 98 µg/m³.     

A real-time prediction method for tunnel boring machine cutter-head torque using bidirectional long short-term memory networks optimized by multi-algorithm

Based on data from the Jilin Water Diversion Tunnels from the Songhua River (China), an improved and real-time prediction method optimized by multi-algorithm for tunnel boring machine (TBM) cutter-head torque is presented. Firstly, a function excluding invalid and abnormal data is established to distinguish TBM operating state, and a feature selection method based on the SelectKBest algorithm is proposed. Accordingly, ten features that are most closely related to the cutter-head torque are selected as input variables, which, in descending order of influence, include the sum of motor torque, cutter-head power, sum of motor power, sum of motor current, advance rate, cutter-head pressure, total thrust force, penetration rate, cutter-head rotational velocity, and field penetration index. Secondly, a real-time cutter-head torque prediction model's structure is developed, based on the bidirectional long short-term memory (BLSTM) network integrating the dropout algorithm to prevent overfitting. Then, an algorithm to optimize hyperparameters of model based on Bayesian and cross-validation is proposed. Early stopping and checkpoint algorithms are integrated to optimize the training process. Finally, a BLSTM-based real-time cutter-head torque prediction model is developed, which fully utilizes the previous time-series tunneling information. The mean absolute percentage error (MAPE) of the model in the verification section is 7.3%, implying that the presented model is suitable for real-time cutter-head torque prediction. Furthermore, an incremental learning method based on the above base model is introduced to improve the adaptability of the model during the TBM tunneling. Comparison of the prediction performance between the base and incremental learning models in the same tunneling section shows that: (1) the MAPE of the predicted results of the BLSTM-based real-time cutter-head torque prediction model remains below 10%, and both the coefficient of determination (R²) and correlation coefficient (r) between measured and predicted values exceed 0.95; and (2) the incremental learning method is suitable for real-time cutter-head torque prediction and can effectively improve the prediction accuracy and generalization capacity of the model during the excavation process.     

Application of non-linear regression analysis and artificial intelligence algorithms for performance prediction of hard rock TBMs

Prediction of machine performance is an essential step for planning, cost estimation and selection of excavation method to assure success of tunneling operation by hard rock TBMs. Penetration rate is a principal measure of TBM performance and is used to evaluate the feasibility of using a machine in a given ground condition and to predict TBM advance rate. In this study, a database of TBM field performance from two hard rock tunneling projects in Iran including Zagros lot 1B and 2 for a total length of 14.3 km has been used to assess applicability of various analysis methods for developing reliable predictive models. The first method used for this purpose was principal component analysis (PCA) which resulted in development of a set of new empirical equations. Also, two Soft computing techniques including adaptive neuro-fuzzy inference system (ANFIS) and support vector regression (SVR) have been employed for this purpose. As statistical indices, root mean square error (RMSE), correlation coefficient (R²), variance account for (VAF), and mean absolute percentage error (MAPE) were used to evaluate the efficiency of the developed artificial intelligence models for TBM performance prediction. The results of the analysis show that AI based methods can effectively be implemented for prediction of TBM performance. Moreover, it was concluded that performance of the SVR model is better than the ANFIS model. A high correlation was observed between predicted and measured TBM performance for the SVR model. This study shows the feasibility of using these systems and subsequent work is underway to expand the database of TBM field performance and use the aforementioned methods to develop a more comprehensive TBM performance prediction model.     

Evaluating D&B and TBM tunnelling using NTNU prediction models

Drill and Blast (D&B) and Tunnel Boring Machine (TBM) are the two dominating excavation methods in hard rock tunnelling. Selection of the cost effective excavation method for a tunnel is a function of tunnel cross section area, rock conditions, tunnel length, availability of skilled labour and proper equipment, and project schedule. Over the past few decades, major technological development and technical advances have been achieved in both methods. Yet, in many tunnelling projects, choosing the excavation method is still a challenge and requires considering pros and cons of each method and estimating construction time, costs, as well as post construction and operation & maintenance, and related risk in the planning phase. In this study, the productivity and efficiency of the D&B and TBM options for excavating certain size tunnels have been examined. The analysis is based on recent NTNU prediction models for advance rate and unit excavation cost for given ground conditions and tunnel geometry. For excavation of large size tunnels in very hard rock, the D&B method seems to be the cost effective choice irrespective to tunnel geometry. This is compared to smaller long tunnels with good boreability were the TBM has higher advance rate. The tunnel size and rock conditions have higher impact on the TBM performance and costs than for D&B. This refers to lower risk of using D&B where the use of this method is otherwise justified.