Abstract Unconfined compressive strength (UCS) of rocks is one of the most important parameters in rock engineering, engineering geology, and mining projects. The accuracy of laboratory test results relies highly on the quality of specimens. In some situations, obtaining good quality specimens is difficult, time-consuming and expensive. In view of this, research on the development of predictive models to determine the UCS of rocks has been carried out for more than three decades. In this paper, an intelligent approach applying the Mamdani fuzzy model was performed to predict the UCS of sedimentary rocks from a Singapore underground excavation project. For the prediction of UCS, parameters including bulk density, porosity and point load test results obtained by laboratory tests on rock cores were used as input parameters. Those predicted UCS results were compared with the results of statistical models using determination of the coefficient of correlation (R 2 ) and root mean square error (RMSE). The comparison reveals that the performance of the fuzzy model is better than the statistical models. This approach is recommended to provide preliminary estimates of UCS for similar sedimentary rocks in other parts of Singapore. As it is the first attempt of such with limited data, it is advisable to update the current model when additional UCS results are available in local civil and mining engineering projects. 1. Introduction The stability of any excavated structure in rock mining project depends on many geomechanical properties of intact and rock mass such as unconfined compressive strength (UCS) and tensile strength. Bieniawski (1974) concluded that UCS is required more often than any other rock properties since this parameter is essential for the stability of underground excavations, access tunnels, roadways, etc. There are standard procedures for measuring the UCS of rock specimens (ISRM, 2014 and ASTM D7012, 2014). These conventional procedures included the fundamental laboratory tests and correlated index tests. In the fundamental laboratory test, the deformation of a rock specimen is measured while the unconfined compressive force is increased. The stress at which the specimen failure occurs is taken as the peak strength. The second approach uses index tests to calculate the UCS instead of directly measuring it. The main advantages of the use of index tests are low costs involved and the measuring equipment is portable (Meulenkamp and Alvarez, 1999). However, in order to carry out these standard tests, good quality specimens such as cylindrical core or block specimens are necessary to be prepared with high accuracy. However, it is sometimes impossible to obtain good quality specimens in the weak or highly fractured rock mass (Alber and Kahraman, 2009).

Abstract Mine development and operation in North and South America as well as Australia and Asia manage rockbursting conditions as a routine but always complex and difficult challenge. The hazard and the associated risks can be managed based on local experience, monitoring, informed data rich modelling and holistic understanding in a mature mining operation. In new mines, including large scale caving operations however, rockbursting poses a significant hazard to billions of dollars in initial infrastructure that must be designed and constructed prior to mining with minimal data and no local experience. In deep tunnelling around the world, a similarly blind approach (tunnelling into previously unexcavated ground) to hazard prediction and risk management must be taken. Blind development for large infrastructure and mine access tunnelling is being carried out around the world at depths in excess of 2km. In complex tectonic regions, tunnels below 500m can experience rockbursting due to high horizontal stresses. In addition, while many of the empirical and semi-empirical tools for the prediction of burst prone ground are based on homogenous and massive rock conditions, tunnels in alpine areas and tunnels built for mine or mine panel access are, by nature, in highly heterogeneous and structured rock (joints, veins, fabric). The author focusses in this paper on the prediction of brittle response potential to excavation under high stress and on the associated but separate potential for rockbursting within a tunnel or mine development heading. The rockburst mechanism considered here is tunnel generated dynamic rupture or strain bursting, distinct from remote or mine-generated events impacting the tunnel excavation. Consideration of rock petrology, fabric, and mechanical parameters allow an initial estimate of brittle response. The potential for energy storage and rapid release must be accounted for to understand the burst potential for as-designed tunnels. 1. Introduction Rockbursts are explosive failures of rock which occur when high stress concentrations are induced around underground openings (Hoek 2006) in brittle rock or rockmasses with brittle structure. In mining, there are many different mechanisms that lead to rockbursts. Pillar failure can be very violent if the pillar core reaches its capacity and the mine geometry is such that instantaneous deformations (system unloading) are large. Large stress changes associated with large scale mining can result in fault slip distant from the drift or shaft but capable of inducing sympathetic strain bursts (due to stress wave propagation) or seismically induced ground falls. In tunnelling, however, the most important mechanism is strain bursting of walls and the tunnel face, with or without structural control and as a result of the complex stress path within the near-field rock as the tunnel advances (Diederichs et al 2013).

Abstract The authors developed a system for predicting geological conditions not just of excavated areas but also ahead of the tunnel face by processing drilling data using Ordinary kriging, a geostatistical approach, and by visualizing conditions in real time. Using auto-controlled face drilling rigs, the authors confirmed the easy and rapid acquisition of drilling data from blast holes and rock bolt holes, along with three-dimensional coordinates. The authors deployed the system in real time to decide whether pre-supports such as forepoling and facebolts would be necessary for a strongly sheared slate area at tunnel faces of the Shin-Kuzakai Tunnel (provisional name). The prediction provided information essential for decisions on construction methods. It also enabled safe streamlining of the process of deploying pre-supports. The authors also compared the applied support systems with P-wave velocity distributions of the rock mass derived from specific energy of drilling and rock mass classification distributions. 1. Introduction Choosing appropriate support systems and pre-supports are the key to ensuring quality, stability, and safety in tunnel construction projects. Typically, the tunnel face is observed during the excavation to evaluate geological conditions. The tunnel face is generally observed once a day, while excavations usually occur around four times a day. This means changes in geological conditions may be overlooked. In addition, as demonstrated by past incidents, the presence of a weak layer behind a tunnel face or a side wall may result in the collapse or significant deformation of a tunnel face or wall. With respect to the geological conditions behind a side wall, hardness is determined empirically from drilling data for rock bolt holes. However, acquiring and analyzing this data requires a great deal of time and labor. Obtaining information on geological conditions that would allow real-time decisions on construction plans tends to pose extreme difficulties. Using auto-controlled face drilling rigs whose use is increasingly common, we found a way to quickly and easily acquire drilling data from blast and rock bolt holes. We established a system that predicts geological conditions not just in excavated areas, but ahead of the tunnel face; this method visualizes conditions in real time by processing drilling data based on a geostatistical approach called Ordinary kriging. We applied this system to the construction site for the Shin-Kuzakai Tunnel (provisional name) on the Miyako-Morioka Cross Road and confirmed that the system provides essential information for construction plan decisions.

Abstract While mechanized mining operations are integral part of mining activities in soft rocks, drill and blast fragmentation method still dominates the hard rock mining operations. Mechanized fragmentation technology for hard rock mining should be robust, relatively small, and flexible; however, development of such system has been a challenge due to energy requirement of cutting hard rocks and the short life span of the cutters. These limitations strongly suggest the need for alternative solutions to reduce the energy requirement of hard rock failure. Actuated Disc Cutting (ADC) is a new dynamic cutting method, which uses disk-shape cutters attacking the rock in an undercutting mechanism. This new system dynamically actuates the cutter, while it is moved across the rock. Breaking the rock under direct tension, ADC consumes less energy for fragmenting rock than conventional methods, hence, reducing the overall power requirement of mechanical hard rock excavation. Introducing the characteristics of an ADC system, this paper summarizes the findings of an experimental study, which investigates the rock failure mechanism and the forces associated with the cutting process using a laboratory ADC test unit. 1. Introduction Application of mechanized cutting/fragmentation technologies in hard rock mining environment is impeded with the limitations of the existing technologies: being power, size, mobility, and cutter life. Overall there are two main classes of cutting tools: drag cutters, such as picks, which are used on road header type machines; and indenters, such as roller disc cutters, used on tunnel boring machines. Drag cutters are more efficient tools than indenters by directly encouraging tensile cracks when moved across the rock surface at a certain penetration depth. Indenter, however, are more wear resistant, due to the rolling mechanism of the cutter, while requiring higher forces for breaking the rock under compression induced tensile failure. Susceptibility to wear and failure has therefore limited the use of drag cutters to excavating low-to-medium strength and non-abrasive rocks (Ramezanzadeh and Hood, 2010); indenters, on the other hand, despite their need for very high thrust forces, have found widespread use in full-faced hard rock cutting of civil applications, where large and heavy machines can be used.

Abstract When considering the sequestration of the energy byproducts of radioactive wastes and anthropogenic CO 2 and the efficient recovery of subsurface energy, it is of significant importance to examine the flow and the transport behavior in fractured rocks. In particular, the fluid flow within low-permeability rock masses is often dominated by transport in through-cutting fractures, and may cause hydraulic weakness. Changes in the ambient stress and temperature conditions should affect the transport characteristics of these conduits through combined mechanical and chemical interaction. We have been seeking a consistent understanding of coupled thermal, hydraulic, mechanical, and chemical controls on the hydraulic and transport properties of natural fractures and fractured rocks. In this lecture, the results are presented of water flow-through experiments in rock fractures under various confining pressure and temperature conditions. The experiments follow the progress of the fracture permeability mediated by the coupled processes. Conceptual models are also presented to replicate the experimental measurements by accounting for the chemo-mechanical processes such as the pressure solution. Furthermore, coupled THMC numerical models are shown to predict the long-term change in rock permeability. 1. Introduction Understanding the fluid flow behavior in the deep subsurface is essential for evaluating the performance of many rock engineering projects, such as the geological disposal of high-level radioactive waste and anthropogenic CO 2 , and the enhanced geothermal system (EGS). In the deep subsurface, the fluid flow behavior often depends on the hydraulic properties of the rock fractures (i.e., permeability and aperture) and the spatial distribution of the fracture network. It is well known that the permeability of fractured rocks is influenced by the coupled thermal, hydraulic, mechanical, and chemical (THMC) processes under the deep geological conditions. Notably, under elevated temperature and stress conditions, mechanically and chemically mediated processes may alter the pore structure in rocks, resulting in an irreversible evolution of the flow and transport behavior. An augmentation in permeability may result from the mechanical dilation brought about by shearing and/or mineral dissolution within the pore spaces, while a reduction in it may result from reversible or irreversible mechanical compaction, mineral dissolution at the contacting areas, and/or the clogging of pore spaces by the precipitation of secondary minerals. Reaction-induced mechanism leading to fracture compaction is well-known as pressure solution (e.g., Weyl, 1959; Rutter, 1976) that incorporates three serial processes; mineral dissolution at stressed contacts, diffusive transport of this material along the interfacial thin water film, and ultimate deposition of the mineral matter at the pore wall. This typically results in the loss of porosity, and a reduction in permeability. Many phenomenological and theoretical models have been also developed to describe the progress of this mechanically- and chemically-dependent behavior.

Abstract In classifying the rock slope, it is important to determine the characteristic controlling instability. The difficulties remain in deciding when too many variables in such a complex material. The objective of this paper is to identify parameters that control instability in the rock slope, which has been based on the extensive parametric study of rock slopes. Modelling was carried out using the Universal Discrete Element Code (UDEC), and the slopes are modelled in terms of the percent of strain. Then, the analytical results from UDEC were analysed using a statistical method called Analysis of Variance (ANOVA). This allowed a robust method of identifying the controlling factors governing rock slope stability. From this study, it was found that discontinuity orientation was the most important parameter. This followed by water, slope angle, Volumetric Joint Count (J v ), Joint Roughness Coefficient (JRC), block shape and Joint Compressive Strength (JCS). The explanation is lies in the low stress levels in slope applications, as the blocks are free to move. The movement is along the inclined discontinuity surface, and the JCS is least significant since that the overriding of asperities is more common than shearing of the asperities. However, it should be noted that for JCS, even though it is the least significant factor, the parameter is still significant as indicated by ANOVA analysis.

Abstract This study reports experimental results of not only thermal conductivity but also compression and shear wave velocities for wide range of rock samples recovered in Korea. Total 45 rock specimens are gathered to represent the different origins, mineralogy, and density. The divided bar method is implemented to obtain thermal conductivity in steady-state and piezo-transducers at kHz ranges are used to measure wave velocities. Measured values are subjected to multiple-regression and statistical analysis with dominant factors of density, which allows constructing the correlative relationship between thermal conductivity and stiffness. Artificial Neural Network (ANN) which learns relationships between data predicts thermal conductivity of rock based on collected physical properties using nonlinear multiple-regression. The statistical approaches using optimal interpolation help understanding and extending correlated multiple geophysical properties in characterization of rocks.

Abstract TBM excavation is very common these days and its market is expected to grow at 4–6% every year. In TBM excavation, estimation of cutting performance is of great importance in design stage as well as during construction. The performance is highly dependent on the geological conditions, i.e. characteristics of rock and discontinuities, and operational conditions, i.e. selection of cutter, cutting forces, cutter spacing, etc. For performance estimation, prediction models can be used in which index properties of rock and rock cutting database are correlated. Establishing the database surely takes much time and effort. And the models are subject to be modified when new sets of database are added. Other way of performance estimation is conducting a full scale linear cutting machine test. In the test, both rock and operational conditions can be easily controlled. However, linear cutting machine test is not often carried out because it requires large block of specimen, large stiff frame and precise control system. Therefore, numerical simulation can have benefits over experimental method and it recently provided decent outputs. In this presentation, recent researches on performance assessment of rock cutting including punch penetration test, full size linear cutting machine test, and numerical simulation are introduced.

Abstract In Japan, it is necessary to excavate many hard rock tunnels using a heavy weight (200–300kW class) axial cutter head type roadheaders, which has reason for various conditions of saving the environment from the construction site and so on. And the performance of a heavy weight axial cutter head type road headers is a critical issue in assessing technical and economic feasibility of its application in many tunneling projects. Therefore, we has established a database of heavy weight transverse and axial cutter head type roadheaders performance, which based on the interview for engineer's of many tunnel construction projects. This database contains field data from the tunnel construction sites all over Japan, includes a variety of heavy weight roadheaders of different geotechnical conditions and different cutter head type. And, this database refers to accurate estimation of the production rate, machine utilization, and picks consumption for different geological units to be encountered on the tunnel project planning. This paper presents and discuss the performance model of instantaneous cutting rate (ICR) and pick consumption from this database, especially for excavated of hard rock tunnel by heavy weight (200–300kW class), axial cutter head type roadheaders.

Abstract The Norimoto Tunnels are twin tunnels about 750 m in length in Shin Tomei Expressways that is under construction by the Central Nippon Expressway Company at Shinshiro City in Aichi prefecture. An existing aqueduct tunnel was constructed in 1960 directly below Norimoto Tunnel at about 17 m offset. Before excavation of the Norimoto Tunnels, there were concerns regarding impact on the existing tunnel, both dynamic impact by vibration associated with blasting and static impact by load of loosened rock mass, and displacement associated with tunnel excavation. In this paper, specific actions taken during excavation of the section crossing the existing tunnel and the results will be discussed. Before actual excavation of the section crossing the existing tunnel, a section 208 m in length was set up for monitoring both dynamic and static impact and the impact on the existing tunnel was estimated in advance. To improve the accuracy of numerical analysis of the impact prediction, a test excavation section was set up prior to excavation of the crossing section. The results of numerical analysis were verified based on the measured ground displacement during the test excavation to improve the accuracy of prediction analysis. Excavation of the crossing section was continuously monitored automatically using an embedded 3-directional compound vibrometer for dynamic impact and 3-dimensional extensometer for static impact. As a result of the excavation of the crossing section, since both dynamic and static impacts were observed to correspond approximately to the prior predicted value, the impact on the existing tunnel was reduced to the minimum. Furthermore, no deformation due to the tunnel excavation was observed during visual inspection of the existing aqueduct tunnel. The specific actions taken during this tunnel excavation will certainly contribute to future similar neighboring construction projects.

Abstract Water and mud inrush will cause severe security risk and great economic loss to tunnel construction, it will also cause unrecoverable damage to relevant environment. However, the prediction and prevention still remain as a very difficult problem. In contrast, mud inrush is more harmful, but on which research work is far from adequate. According to geological survey, Lingjiao tunnel is judged to be safe from danger of large-scale water and mud inrush. Unfortunately, a large mud outburst accident happened with bursting mud of more than 40,000 m3, leaving a huge collapse pit on the upper surface of the crossed mountain. Moreover, another mud inrush took place right in the paralleling tunnel nearby one year after the first accident, which triggered another follow-up collapse in the same site as the first time, though it has been reinforced with concrete cover. Facts above show that the two-track tunnel crossed the same mud-filled structure, which was not well-understood. In this study, the mud-inrush process, causes and treatment measures are expatiated. Through case study, we hope to provide reference for prediction and prevention of mud-inrush disaster during tunnel construction in the future.

Abstract To minimize the social cost and interference with existing facilities, tunnels for cable and electricity have been mostly constructed using trenchless technology in recent years. In this study, a new small-scale excavation method using a multi-hammer tunnelling machine was suggested. To evaluate the optimum operating conditions of the machine, linear percussion testing system to experiment with various conditions was established. Linear percussion tests were conducted considering the air pressure and RPM (i.e., two dominant-field operating conditions). The optimum operating condition of a multi-hammer tunnelling machine for three rock classes was obtained from the test results. Based on the penetration rate data measured from a field test, a prediction chart of the penetration rate was proposed for various rock classes.

Abstract Physical model simulations have been performed to determine the effects of underground opening configurations on surface subsidence under super-critical conditions. This paper indicates the importance of the main factors that control the extent of surface subsidence and determines the effects of geometry of underground openings on the angle of draw, the maximum subsidence and the volume of the subsidence trough. A trap door apparatus has been fabricated to perform the scaled-down simulations of surface subsidence. Gravel is used to represent the overburden in order to exhibit a cohesive frictional behavior. In plan view the excavation dimensions are sufficient to induce maximum possible subsidence. The findings can be used to evaluate the subsidence profile for tunnels and caverns in soft ground. The results show that the angle of draw and the maximum subsidence are controlled by the width (W) , length (L) , height (H) and depth (Z) of the underground openings. The width of the subsidence trough can be represented by sets of empirical relations. The relation between opening depth and subsidence trough developed by Rankin is in good agreement with most physical model results for deep openings ( Z/W = 2, 3 and 4). For shallower openings (( Z/W = 1), the Peck estimate is better than Rankin's. The volume of the subsidence trough is largest for ( Z/W = 2.5 and for ( H/W = 0.6, and is about 60% of volume of the underlying opening.

Abstract Mega wedges represent adverse conditions during the excavation work of large underground caverns for storing hydrocarbons. Even in "Very Good" rock (assessed using an empirical approach such as the Q system or the RMR), the potentiality of wedges involving atop heading and one or several benches is a recurrent problem that cannot be neglected in unlined rock caverns, only stabilized using rock bolts. The empirical approaches have not been developed to define the necessary support in this very specific context which requires detecting and stabilizing such very large wedges (mega wedges)using a deterministic approach, based on structural geology. Although minimization of the risk and impact of mega wedges is considered by the cavern basic design and the adaptation of the layout and support, a deterministic re-assessment of the areas where mega wedges may occur shall be done during the excavation phase at a more precise scale, on realistic structural conditions. Established on real cases in granite, this paper presents a practical and balanced methodology that is used from the design to the excavation and support phases. For that, as a first step, structural geology aims at identifying location, geometry and potentiality of mega wedges and then rock mechanics helps to characterize the shearing capability of the identified mega wedge and its putative failure. In addition to the support recommendation, the phasing of the support installation with respect to the excavation work to achieve stable condition of the mega wedges is discussed. Results are promising in term of applicability, even if a better estimation of the shear strength parameters of the discontinuities involved in the failure mechanism can be necessary for local fine tuning.

Abstract Earth-fill structures such as embankments, which are constructed for the preservation of land and infrastructure, show significant amount of settlement during and after construction in lowland areas. The long term settlement of those structures is often measured. In this paper, we examined the applicability of a neural network model for settlement prediction using measurements in the early stage after construction. Simulations using a basic network model showed that when the measurement data used for teaching the neural network accumulated, the prediction was in good agreement with the measurement data, and the variance of predicted values was low. However, the basic model could not predict the settlement behaviour precisely, when the amount of teach data was limited as would be in the early stage after construction. Some improvement was necessary for this model to conduct early settlement prediction. To achieve a higher accuracy in long term settlement prediction from early stage measurements, several improvements to the model are proposed, which generate additional data points and improve the prediction accuracy. Firstly, a cubic spline interpolation technique is used to generate additional data between measurements and regulate the input to constant time intervals. Secondly, statistical techniques are used to weed out predicted data points that are outside preset parameters. Short-term predicted values that satisfy with clear statistical criteria (low co-variance or low standard deviation) are added to the network teach data. The improved network model simulations showed that the accuracy of settlement prediction based on early stage measurements improved significantly.

Abstract In general, discussions concerning noise problems due to blasting excavation of the tunnel focus on the noise caused by the sound propagated by air vibrations. However, sometimes the noise that the ground vibration causes might become a problem. This noise is called a solid-borne sound. When the solid-borne sound caused by blasting vibration occurs inside a house, acoustic radiation takes place at the floor, walls, and ceiling of the living space as the house vibrates. The intensity of the sound can increase depending on the dominant frequency of vibration. It is important to note that especially when the ground is hard and has little discontinuity, the noise is generated by the unrecognizable ground vibration because the vibration is transmitted far away without the high frequency components attenuating. In this study, we report the case of the acoustic radiation (solid-borne sound) caused by blasting vibration during tunnel construction. We compared the measured values with the values predicted by equations of vibration and noise inside and outside of the houses to be preserved. As a result of our study, we found that the noise level inside the house was often larger than just outside the house. The difference between the noise inside the house and that in field can theoretically differ by 10 dB at maximum. When the noise problem in the tunnel excavation is discussed, this result indicates that not only air propagation but also the solid-borne sound should be considered. If the sound measurement in the field is employed as the reference, we need to use control values at least 15 dB lower than the conventional control value.

SYNOPSIS Spalling is the common mode of failure in deep tunnels excavated in strong rocks which results in notches around the tunnel boundary. Recent developments in the analysis of spalling failure have resulted in the empirical Hoek-Brown spalling criterion which is used for prediction of spalling profile. However, this criterion is commonly applied following elastic numerical analyses and in an implicit way. In this paper, an explicit solution is derived for spalling profile around circular tunnels. The stress distribution around tunnel is obtained using the well known analytical elastic solution, and then the locus of points which satisfy the empirical Hoek-Brown spalling criterion is determined. The derived polar function is used for graphical illustration of the effect of governing parameters on the spalling profile. Finally, the results obtained from the solution are compared with observations made in real tunnels. Showing good agreement with in-situ measurements as well as having explicit and easy-to-use form, the derived solution is expected to be extensively used in the future for prediction of spalling profile around circular tunnels.

ABSTRACT Under the observational design and construction method, original tunnel design is immediately verified or modified based on the prediction results of geological conditions from in-situ data sequentially obtained during construction, such as seismic profiling, probe hole drilling and geological ratings at tunnel faces. Because data used and its processing method for geotechnical prediction change according to purposes, this study proposes the integrated prediction system for geological conditions ahead of tunnel faces, which is composed of long-interval, middle-interval, and short-interval prediction subsystems. The long-interval prediction subsystem is used for master planning, such as estimation of the total period and cost of a project. The middle-interval prediction subsystem is used for detailed planning, such as judgment of execution of preliminary reinforcement and selection of tunnel support pattern. The short-interval prediction subsystem is used for final decision making, such as confirmation or modification of the detailed plan and determination of tunnel excavation parameters. The applicability of those subsystems is verified with actual field data obtained in the motorway tunnel project, where Cretaceous granite occurs, involving a main tunnel excavated by the drill and blast method, and an accompanied evacuation tunnel excavated by a TBM in advance of the main tunnel.

ABSTRACT In this study, a series of permeability experiments on both intact rock and a single fracture in sandstone and on a single fracture in siliceous mudstone has been conducted under confining pressures of 3 – 15 MPa, and at temperatures of 20 – 90°C for several hundred days in each experiment. Spontaneous alteration of permeability and dissolved mass fluxes have been observed. In sandstone, the permeability of intact rocks little changed until a couple of hundred days, then it started to increase with time. In mudstone, the permeability in a single fracture monotonically decreased with time and reached a quasi-steady state within one month. However, it started to increase with time after ~100 days. This augmentation in the permeability in sandstone and mudstone should be attributed to mineral dissolution within the void spaces. In contrast with the sandstone results, the fracture permeability in mudstone resumed decreasing with time throughout the rest of the experiment. A coupled chemo-mechano conceptual model, accounting for pressure and free-face dissolutions, is utilizedto follow the evolution of the permeability observed in the experiments. The model predictions show a relatively fair agreement with the experimental measurements, although further considerations and modifications are found to be required.

ABSTRACT Exploring rock mass and machine interaction is not a straightforward task due to complexity and large variability of ground, geology and rock mass conditions even in short intervals. Researchers, however, rarely employ non-linear models to examine the determinants and make little effort to identify a superior prediction model. The main objective of this paper is to compare two TBM performance estimation models with especial task of determining Rate of Penetration (ROP), using Artificial Neural Network (ANN) and multiple linear regression. These approaches were applied to a database compiled from 121 tunnel sections along the Milyang hydro-tunnel for correlating ROP with basic RMR input parameters. Nonetheless, the poor relationship between ground water condition and ROP resulted in exclusion of this parameter from the further analysis and Uniaxial Compressive Strength (UCS), RQD, joint spacing and joint condition were used. The accuracy of developed model was discussed by some statistical measures including correlation coefficient (R), the Root Mean Squared Error (RMSE), the accuracy factor (A f ), and the bias. Results of this study showed that ANN model has lower RMSE values for in-sample and out-of-sample forecasting. This indicates that the non-linear ANN model generates a better ft and predict of the panel data set than the regression model, and ANN is capable of catching sophisticated non-linear integrating effects in TBM performance problem.