Search

Underground construction has become a necessity considering social, technological, economical, and sustainable advancements in society. Tunnel boring machines (TBMs) can be considered as an efficient method of tunneling because of its high productivity and applicability in a variety of subsurface conditions. The performance of a rock TBM is assessed by its advance rate that is a function of penetration rate and utilization factor. While there are various existing models that can reasonably predict the penetration rate, only few models have the capability to estimate the utilization, and the accuracy of models predicting the utilization is highly dependent on the incorporation of complexities involved in different tunneling activities. The geology-related delays include those incurred for ground stabilization, including installation of rock bolt, wiremesh, strap, or channel, placing of ribs and lagging, gripping process, control of water inflow, cutter change and inspection, probe drilling, scaling, muck removal, and foam or grout injection for ground stabilization. This study aims at determining the most critical ground support activities affecting the TBM performance using discrete event simulation. For this purpose, six projects that utilized open TBMs were analyzed. Comparing the individual ground support measure installations, steel ribs showed the maximum downtime of 7% compared to water inflow at 4.8% and cutter change at 2%. The results show that geology-related delays indeed contribute significantly to the variations in utilization factor.

ABSTRACT: One approach to improve the reliability of predicting the response of tunnels is the "design as you go" method, where tunnel response is monitored during tunneling using field instrumentation. In the present study, various machine learning (ML) techniques are evaluated, and the best approach is proposed for the back-analysis of tunnel response. Incorporating in-situ stress conditions in back analysis is also essential as these are responsible for short- and long-term deformations. The study aims at predicting the cohesion and friction angle for a tunnel by setting the monitoring data and in-situ stress conditions. The ML techniques are combined with the results obtained from the commercially available finite difference code FLAC, widely used in geotechnical modeling. The ML algorithm with maximum accuracy is then used to determine the input parameters using the synthetic monitoring data. The present study evaluates a single tunnel with an overburden of 1000 m and an in-situ stress ratio ranging from 0.25 to 3 to prepare a training and validation set. The results show that with an acceptable 0.1% error, the Levenberg-Marquardt Back-Propagation Method (LMBP) of Artificial Neural Network (ANN) is found to be the most suitable ML algorithm. 1. INTRODUCTION The key to the meticulous design of the underground structure is the careful assessment of ground parameters. Evaluation of rock mass behavior is crucial in determining the necessary support measure design that ultimately defines the safety of underground structures. This assessment is based on the results obtained from the numerical models with rock mass parameters as input data prior to construction. Bertuzzi (2017) pointed out that the elastic modulus of rock mass is often calculated using the correlations obtained from the rock mass classification system. Many researchers have expressed difficulty in interpreting the rock mass parameters using the in-situ testing as high variability in the results can be obtained, and the average behavior can be assessed by performing a large number of tests (Vibert and Ianos, 2015). Back-analysis of the rock mass parameters using the monitoring data is another approach that has proven to be helpful in optimizing underground structures’ overall design and is the aim of the present study. The back-analysis aims to validate the model parameters and improve the forward modeling of subsequent excavations. Re-calibration of the back-calculated parameters optimizes the design and overall project. Callisto and Ricci (2019) studied the tunnel damage response under seismic loading. It was found that back analysis contributed to assessing the primary causes of damage: geometry and dynamic response of lining. The crucial point for performing the back analysis is identifying the parameters that affect its accuracy.

Improving the accuracy of models for tunnel boring machine (TBM) performance prediction and estimation of utilization factor is the focus of many ongoing studies in mechanized tunneling. Utilization factor is controlled by tunnel geology, rock mass condition, and size of the tunnel as well as tunneling activities such as maintenance, utility installation, transportation, surveying, unexpected breakdowns and miscellaneous downtimes. In this study, modeling tunneling activities and downtimes using discrete event simulation approach is used to predict TBM utilization factor and advance rate. Two mechanized tunneling projects were considered for the development of the model and verification of the results. A database was developed using Karaj Water tunnel project to generate the time distributions for various tunneling activities required as input for simulation model. Arena© software was used for the simulation of TBM operation in this project. The result of modeling was used to simulate Nowsood tunnel project for verification of the concept and discrete event simulation approach. The results showed a good correlation between the predicted TBM performance parameters and observed machine utilization. This exercise showed that the model is capable of predicting utilization factor based on available data for the ground conditions, operational settings, and tunneling activities.

The present work aims to understand the stress–strain response of a metamorphic rock, and gneiss under high loading rate for different specimen diameters and slenderness ratios through detailed tests. The high strain rate characterisation of gneiss rock is done for two different diameters and five different slenderness ratios of the rock specimens using a 76 mm-diameter split Hopkinson pressure bar (SHPB) device in an effort to understand the standard specimen dimension for gneiss in SHPB test. The stress–strain response of the rock specimens is studied by varying the length of the striker bars and the gas gun pressure values, systematically. The petrological and static characterisation of the gneiss rock is also carried out to assess the response of the rock specimens. Finally, a methodology is proposed to characterize gneiss rock specimens under high loading rate. Furthermore, numerical simulation of SHPB test on gneiss rock is performed using strain rate-dependent Johnson–Holmquist (JH-2) model available in the finite-element software package, LS-DYNA. The simulation results are compared with the experimental data, and thus, the parameters of JH-2 model for gneiss rock are determined.

In the present work, dynamic stress–strain response of five sedimentary and three metamorphic rocks from different regions of India, e.g. Kota sandstone, Dholpur sandstone, Kota limestone, Himalayan limestone, dolomite, quartzite, quartzitic gneiss and phyllite have been investigated through split Hopkinson pressure bar test at different strain rates. The dry density, specific gravity, static compressive strength and tensile strength values of the rocks have also been determined. Petrological studies of the rocks have been carried out through X-ray diffraction test and scanning electron microscope test. It is observed from the stress–strain response of the rocks that the peak stress increases with increasing strain rate. Dynamic increase factors for the strength of these rocks have been determined by comparing the dynamic and the static peak compressive stresses and correlation equations are proposed.

The fundamental components in predicting the TBM performance (advance rate) are the rate of penetration and utilization. Extensive studies have been conducted to accurately estimate these parameters. Inclusion of various tunneling activities and their interdependencies, the impact of the site set up and management, and the role of human factors on the operation make an estimation of utilization a complex task. A few models have been developed to predict machine utilization, and these models are based on a statistical analysis of available TBM project databases. Apart from the existing models, the use of discrete event simulation (DES) for estimation has also been examined for TBM performance prediction. The sensitivity and accuracy of all these prediction models demonstrate a strong influence of geological conditions and site settings. The present study compares outcome predictions from six empirical models and DES simulations for eight projects. The variability of the predictions is assessed, and the prediction outcomes are tested for “validation” by comparison with results from the recorded observation at the tunneling sites. The result showed that the difference between the predicted utilization using discrete-event simulations and the actual value available from site data was 7%. Comparing the utilization obtained from other empirical methods, the considerable deviation can be seen. Hence, it is concluded that the DES approach provides for a more accurate estimation of machine utilization while offering the flexibility to customize the model on the basis of both equipment selection and the availability of data to incorporate the complex interdependencies of different tunneling activities.

The present study aims to investigate the physical, petrological, and mechanical properties of ten different types of rocks in the Indian subcontinent, required for the design of structures built in rocks. A comparative assessment is done for the stress-strain response of all the rocks collected under three main categories, e.g., igneous, sedimentary, and metamorphic. The physical properties, e.g., densities and slake durability indices, are determined. The peterological tests are carried out by scanning electron microscope and the X-ray diffraction technique. Mechanical properties, e.g., hardness, uniaxial compressive strength, elastic modulus, and Poisson’s ratio, of the 10 different types of rocks are obtained through static loading at a strain rate of 0.001/s. Lastly, various correlation equations are proposed relating the uniaxial compressive strength, density, elastic modulus, and hardness of the 10 rocks. And the rocks are classified based on the Deere-Miller rock mass classification system.