Skip Nav Destination
Filter
Filter
Filter
Filter
Filter

Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004
Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004
Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004
Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004
Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004

Update search

Filter

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

- Title
- Author
- Author Affiliations
- Full Text
- Abstract
- Keyword
- DOI
- ISBN
- EISBN
- ISSN
- EISSN
- Issue
- Volume
- References
- Paper Number

### NARROW

Format

Subjects

Date

Availability

1-17 of 17

Keywords: machine learning

Close
**Follow your search**

Access your saved searches in your account

Would you like to receive an alert when new items match your search?

*Close Modal*

Sort by

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004

Paper Number: ARMA-04-478

... Artificial Intelligence pressure condition neural network correlation net pressure mean effective stress pore pressure prediction shale sandstone overpressure

**machine****learning**Upstream Oil & Gas Muderong Shale compaction Dewhurst Reservoir Characterization pore pressure...
Abstract

ABSTRACT: A methodology for remote pore pressure prediction using seismic attributes is presented, based on rock physics responses to various overpressuring mechanisms. A series of laboratory acoustic tests were performed on reservoir sandstones and shales, simulating normal compaction, disequilibrium compaction, fluid expansion and tectonic mechanisms of overpressuring. Sandstones showed lower V p /V s ratios during fluid expansion overpressuring as compared to both normal and disequilibrium compaction. For shales on a tectonic stress path, velocities and elastic constants all increase with increasing mean effective stress. P- and S-wave anisotropy are initially high and diverge with increasing mean effective stress. Seismic attributes of these laboratory waveforms were derived to verify which attributes were most sensitive to changing pore pressure conditions. Positive correlations between effective stress and several instantaneous seismic attributes were established allowing direct mapping of seismic attribute changes into definite values of effective stress. This methodology has been tested on a 3D seismic dataset from the Northwest shelf of Australia and shows good agreement with both the distribution of normally pressured and overpressured wells as well as the magnitude of the overpressures present. 1. INTRODUCTION Pre-drill prediction of overpressure is usually achieved through manipulation of seismic data, founded on the empirical relationship between effective stress and seismic velocity [1,2,3]. Velocity-based methods usually require a normal compaction trend and deviations from this normal compaction trend are taken to be indicative of the presence of overpressure. However, velocity-based pore pressure prediction methods are not reliable under all geological conditions. The relationship between stress and seismic P-wave velocity is, in general, non-unique because P-wave velocity is affected by other factors such as lithology and stress history. The reliability of estimated P-wave seismic velocities from surface measurements also decreases with target depth, signal to noise ratio and structural complexity for example. Consequently reliable remote prediction of abnormal pore pressure requires research into alternative approaches. The application of seismic attributes to predict overpressure based on VSP analysis was proposed by [4]. They conducted an analysis on field data and concluded that instantaneous seismic attributes are sensitive to both variations in lithology and pore pressure. Consequently they suggested that sequence attributes may be more relevant for direct detection of overpressured areas, although the methodology they proposed was qualitative in nature. However, our understanding of factors controlling seismic attribute response at the field scale is limited, but this can be enhanced through the derivation of seismic attributes on core samples at ultrasonic frequencies under controlled stress and pore pressure conditions. Previous work has shown that stress path is an important control on ultrasonic response in sandstones [5]. These authors subjected reservoir rocks to pressure conditions simulating normal compaction, disequilibrium compaction and fluid expansion and showed that velocity response and V p /V s ratios were stress path dependent. The full elastic tensor and anisotropy parameters for a shale were derived by [6] under low stress conditions and this work is expanded here to higher stress levels and anisotropic stress states.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004

Paper Number: ARMA-04-426

... well logging

**machine****learning**frequency Artificial Intelligence ISBN imagery cohesion earth material permittivity reflectance normal stress porosity saturation Reservoir Characterization US government friction phreatic surface dielectric permittivity relative permittivity tangent...
Abstract

ABSTRACT: The mechanical properties of an earth material are related to it's dielectric permittivity, or the per-unit dimensional extent of which an electric charge distribution in a material can be polarized or distorted by the application of an electric field. The activation of an electric charge distribution within the material is dependent on the composition of the material, and the electrical resistances of the material's components. Material composition is related to the material's mechanical properties. Dielectric permittivity can be measured through the earth material's absorption of wave energy. The degree of absorption is applied to calculating the material's physical properties such as density, porosity, and also the material's shear strength values at varying depths in terms of cohesion and internal friction angle. Dielectric permittivity is also used to calculate the moisture content and porewater pressures. The level of the phreatic surface is inferred through the permittivity in relation to the degree of absorption of wave energy, the porosity, and the degree of saturation 1. INTRODUCTION Geotechnical data was collected at a copper sulphide tailings basin, primarily through cone penetrometer testing and logging. Field data was gathered in September and October of 1997. Included was data pertaining to tailings density (¿), cohesion (c), angle of internal friction (F), and the depth to the phreatic surface (l). Multispectral imagery of the site was acquired in April of 1999. This imagery had a ground resolution of one meter, and was in 8-bit format. For each test site, pixels were analyzed on the imagery and geotechnical parameters were calculated from the pixel values, which represent the degree of wave signal remission. Results were then compared with the field data. 2. DETERMINATION OF DIELECTRIC PERMITTIVITY The relative dielectric permittivity e r of the material is determined through signal reflectance at varying frequencies. By observing that the earth material to be studied is a dielectric material, and assuming negligible refraction losses of waves hitting earth material, one can ideally assume that reflectance is complementary to absorbance, or [1]: A(f) + R = 1 (1) where A(f) = frequency-dependant attenuation or signal decay (absorption), and R is the signal reflectance (or radiance), and is comprised of wave components: R = .5(R s + R p ) (2) R s and R p are the reflectance/remittance of s- and p-waves of a transmitted wave signal. The value of R can be found, in terms of percent reflectance, by examining the digital number of a single picture element (pixel) of an image, which merely represents the strength of the remitted signal. Most imaging systems use 8-bit binary data formats, which means that pixel values range from 0 (no signal remission) to 2 8 , or 256. The pixel's brightness value, or digital number (DN) will fall within this range for 8-bit data, 256 will be the brightest possible value of any pixel in the image. The dielectric permittivity of the earth material can be found through the amplitudes of the propagation vectors of the material's electric fields, if the reflectances of the s- and p-wave components of the transmitted signal R s and R p can be determined.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS), June 5–9, 2004

Paper Number: ARMA-04-488

... Artificial Intelligence Upstream Oil & Gas exponent aperture outcrop sandstone dolostone aperture size cumulative frequency Cupido dolostone power law Texas coefficient pottsville sandstone intensity

**machine****learning**fracture characterization Thickness kinematic aperture power-law...
Abstract

ABSTRACT: Opening-mode fractures in siliciclastic and carbonate host rocks from different geologic settings consistently show power-law aperture-size distributions when measured along one-dimensional scanlines. Exponents (slopes) and pre-exponential coefficients (intensities) of the power laws, were compared in order to constrain the range of fracture intensities of various aperture sizes over the observation scales. The coefficients range over two orders of magnitude, from 0.1 to 62.4, and the exponents range from -0.45 to -1.36. On a compilation plot of power laws a broad wedge-shaped envelope defines the natural range of frequencies of fractures for aperture sizes from approximately 1000 to 0.01 mm. There is an inverse correlation between the coefficient and exponent for most of the data sets. Power laws having steep slopes tend to have low intensities, whereas power laws having shallow slopes tend to have higher intensities. There are, however, a few exceptions to this correlation. The reasons for the correlation and the departures from it are considered in the light of geologic and mechanical properties at the time of fracture formation for each case. The subcritical crack index, mechanical layer thickness, total strain and strain rate are all likely controls over the relative proportion of narrow to wide fractures in a population. 1. INTRODUCTION Empirical studies have shown that the aperture sizes of opening-mode fractures within a single set are self-organized into power-law distributions [1-3]. Power-law aperture-size distributions ranging over five orders of magnitude have been reported in sandstone and limestone [1] and over three orders of magnitude in dolomites [4]. The range over which power-law distributions apply is typically that observed at the field outcrop scale down to microfractures; on the order of 1 m to 1 µm. Small and large opening-mode fractures that have the same orientation may be different size fractions of the same fracture set. Power-law descriptions of fracture aperture populations take the form: F = ab -c , where F is cumulative fracture frequency, a is the coefficient, b is the fracture aperture, and c is the exponent of the power-law relationship. At any given aperture size, the intensity of fractures is defined by the pre-exponential coefficient, which governs the position of the curves on the ordinate, and the exponent, which is the slope of the curves and is negative. The exponent reflects the relative proportions of wide and narrow fractures in the population. A small exponent (shallow slope) indicates a fracture population with few narrow fractures relative to wide fractures, whereas a large exponent (steep slope) indicates a population with a large number of narrow fractures relative to wide fractures. The values of the exponent and coefficient are different for different fracture sets but the range and reasons for the variation have not been addressed previously. The aim of this contribution is to present a compilation of opening-mode fracture aperture-size data sets so that the aperture-size distributions from different rocks may be compared and the range of intensities of naturally occurring fractures of different aperture sizes may be constrained.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper Number: ARMA-04-517

...

**machine****learning**Upstream Oil & Gas variance random variable gamma pdf Standard Deviation Artificial Intelligence strength failure mass safety factor US government waviness wedge effective normal stress shear strength tensile strength metals & mining Engineering failure...
Abstract

ABSTRACT: The point estimation method can be applied to the safety factor (SF) equation for any specified rock slope failure mode (such as plane shear, step path, or wedge) to obtain reliable estimates of the mean and standard deviation of the SF probability distribution. A gamma probability density function is recommended for modeling this probability distribution, because it allows only for positive values and is flexible enough to provide symmetrical shapes and right-skewed, exponential-type shapes for the SF distribution. The mean and standard deviation define this distribution, which then can be integrated numerically from 0 to 1 to obtain the probability of sliding, P S (portion of the SF distribution where SF<1.0). The overall probability of failure, P F , for the potential slope failure mass is the joint probability that the rock discontinuities are long enough to allow kinematic failure (P L ) and that sliding occurs along the rock discontinuities (P S ); that is, P F = P S P L . This method for estimating the probability of sliding is extremely efficient computationally, and thus, expedites slope stability simulation routines used by NIOSH software to stochastically describe rock slope behavior and assist the engineer in designing catch benches for large rock slopes. Enhanced bench design translates into increased operational efficiency and safer working conditions in open pit mines and quarries. 1. INTRODUCTION Stochastic simulations of fractured rock masses can provide valuable information for the engineering design of rock slopes, particularly when the natural geologic discontinuities may form potential slope failure modes. An essential component of such engineering simulations is being able to compute the probability of sliding for a given potential failure mass once the geometry of that failure mass has been identified through spatial rock-fracture simulations. In cases where several thousand (or ore) simulations of possible failure geometries are needed to provide a realistic representation of the rock slope, computational efficiency is essential for the repetitious protocol used to calculate the probability of sliding. An example of applying this type of geotechnical approach to rock slope design was presented by Miller and others [1] in a paper focused on the design of catch benches for open pit mines and quarries. This issue is important to NIOSH (National Institute for Occupational Safety and Health) as part of its research mission to improve safety and health in the mining industry. Between 1995 and 2003 there were 42 reported fatalities dueto slope failure accidents at surface mines in the United States, at least one and as many as eleven each year. Additional accident statistics collected by the Mine Safety and Health Administration (MSHA) have shown that loose material from slope and bench failures can pose significant safety hazards to miners. To address these concerns the NIOSH Spokane Research Laboratory has been developing and testing rock slope stability software ver the past few years to provide advanced technical tools for analyzing bench stability. A key element of this software is a module used to compute the probability of sliding for a given viable failure geometry that has been simulated for the bench under study.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper Number: ARMA-04-511

... reservoir geomechanics relation

**machine****learning**hydraulic fracturing Artificial Intelligence Reservoir Characterization rock compliance experimental result compressive loading Upstream Oil & Gas compliance compressive stress rock specimen deformation hyperbolic model specimen...
Abstract

ABSTRACT: A new semi-empirical model that predicts fracture deformation under normal compressive loading is presented. The development of a simple exponential model is given first after which a modified and more general exponential model, with an additional degree of freedom in the model parameters, is presented. The simple and the modified exponential models are then compared to available fracture closure models, namely the empirical Barton-Bandis hyperbolic model, and a power-law model based on Hertzian contact theory, to determine how good they fit the results of fracture closure experiments conducted under monotonically increasing normal compressive loading. A new parameter called the half-closure stress, s½, is introduced and is used, in addition to the maximum fracture closure, ¿¿ m , in the model fitting procedures for the Barton-Bandis and the simple and generalized exponential models. The half-closure stress is shown to be related to the initial normal stiffness, K ni , used in the original Barton-Bandis model. An additional parameter, n, is used in fitting the modified exponential model to the experimental data. Of the models presented herein, the modified exponential model was found to provide the best fit to the experimental data, for the same values of s½ and ¿¿ m , over the entire range of compressive stresses. The power-law model based on Hertzian contact theory was found to be unsuitable for accurate prediction of fracture normal deformation behavior. 1. INTRODUCTION Fracture deformability resulting from normal compressive stress is of fundamental importance to the study of the hydraulic and mechanical behavior of rock discontinuities. Fracture deformation directly affects the factors that govern the hydraulic conductivity of single fractures such as aperture distribution, contact area distribution, and spatial connectivity of the apertures [1,2]. Since fracture geometry networks and fluid flow through single fractures govern the hydraulic behavior of fractured rock masses, it follows directly that deformability of single fractures due to the action of compressive stress would affect the hydraulic properties of a rock mass. It is also generally understood that the mechanical behavior of rock masses is controlled significantly by the deformation of discontinuities [3]. The most fundamental properties of the bounding surfaces of a fracture affecting fracture deformation include the rock type, weathered state and matedness of the surfaces, and the spatial and size distributions of asperities on the surfaces. The roughness of each bounding fracture surface is directly related to the size and spatial distributions of the surface asperities, whereas the aperture size and spatial distributions, and the contact area distribution are functions of the cross-correlation of the surface asperity spatial and size distributions. The mechanical strength and deformability of the asperities are functions of the rock type and weathered state of the bounding surfaces. It is also a well known experimental observation that the shear displacement of the bounding surfaces of a fracture, that governs fracture mating, drastically affects the aperture and contact area distributions, even for zero to moderate normal loads, for which the asperity spatial and size distributions of the individual surfaces are practically unchanged [4,5].

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper Number: ARMA-04-519

... real time system Upstream Oil & Gas dimension Artificial Intelligence orientation hydraulic fracturing displacement

**machine****learning**fracture dimension tiltmeter measurement procedure fracture orientation treatment efficiency real-time estimation inversion hydraulic fracture...
Abstract

ABSTRACT: In this paper, we describe a method to estimate, in real-time, the volume of a hydraulic fracture from tiltmeter measurements recorded during the treatment. The resolution of fracture dimensions and orientation is discussed at first. A limit, expressed in terms of the distance between the tiltmeters and the fracture, above which only the volume and orientation of the fracture have an effect on the tilt field is established. The proposed method of analysis takes advantage of this, often neglected, fundamental properties of the elastic kernels (St. Venant principle). We recognize that, in most cases without additional information, only the fracture volume and orientation can be accurately estimated from tiltmeter measurements. The knowledge of the fracture volume estimated from tilt data together with the injected volume at time t furnish the efficiency of the treatment at this time. Furthermore, because the method employed to analyse the tilt data is computationally efficient, the volume and efficiency can be obtained in real time (i.e. as the data arrive ) throughout a hydraulic fracturing treatment. 1. INTRODUCTION Hydraulic fracturing is a technique widely used for the stimulation of oil and gas reservoir (Economides & Nolte, 2000). Fracture orientation must be known to determine well layout and spacing, especially in reservoirs with permeability anisotropy. In some environments, there is considerable uncertainty about whether fractures are growing in horizontal or vertical planes, which can significantly effect the overall size of the fracture formed. Fractures treatment schedule designs which assume a vertical orientation will often fail in execution if the actual fracture is horizontal. Ultimately, the success of a hydraulic fracture treatment depends on placing the correct amount of proppant in the fracture to produce a conductive channel with a size and shape that provides an overall optimal stimulation effect. Fracture olume and efficiency are important in design because these parameters help determine timing and volumes of for pumping clean and slurry fluid stages. In order to "map" hydraulic fractures, several type of indirect measurements are typically used such as micro-seismic acoustic monitoring, tiltmeter mapping, and treatment pressure analysis. In this communication, our interest lies in the analysis of tiltmeter measurements performed during hydraulic fracturing jobs. Despite the now common use of tiltmeters to map hydraulic fractures in the petroleum industry (Wright et al., 1999), it is not precisely clear what information on the fracture can and cannot be obtained from such measurements. Based on practical experience, (Cipolla & Wright, 2000) lists some of the fracture quantities better resolved by surface or borehole tiltmeters. In this paper, we present formal results we have recently obtained regarding the issue of determining the geometry of hydraulic fractures from tiltmeter measurements. We then discuss practical implications of these results. Specifically, we show that due to the elastostatic nature of the problem, the effect of the dimensions of the fracture fades extremely fast spatially such that, in most case, tiltmeters only sample the far-field effect of the hydraulic fracture. This far-field deformation pattern only depends on the orientation of the fracture (anisotropic displacement field) and the volume of the fracture (intensity of the recorded tilts), the two being coupled.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper Number: ARMA-04-521

... be calculated. A high level of redundancy (only six of the eighteen strain readings are required to determine the full stress change tensor) provides a strong measure of the confidence that can be placed in the integrity of the result.

**machine****learning**fracture mechanics stress change data...
Abstract

ABSTRACT : This paper describes the successful measurement of stress changes induced in a crystalline rock mass adjacent to a full size hydraulic fracture. A hydraulic fracture was initially formed using water and subsequently reopened using cross linked gel. The stress changes measured during these treatments are compared with the results of 2-D and axisymmetric models. The full three dimensional stress changes were measured using four ANZI stresscells installed and tested in situ prior to the start of hydraulic fracturing. The instruments were logged at 15 second intervals throughout the hydraulic fracture treatments to provide a time history of the complete three dimensional stress changes that occurred as each hydraulic fracture grew toward and then passed close to the instruments. Analysis of the stress change data provides information about the fracture rate and mode of growth, orientation, and about the excess pressure acting inside the fracture to open it. The stress change data confirms that the hydraulic fractures grew as openingmode rather than shear mode fractures. Stress change calculated using a radial fracture geometry model with the fracture opened by a uniform pressure fit the measured stress change quite well. The fit obtained helps establish the orientation and growth rate for the fractures generated at the site. 1. INTRODUCTION Hydraulic fracturing has long been used to induce fractures in rock masses for reservoir stimulation and other purposes. However, it is rare to have the opportunity to measure the actual stress changes that are caused by a full scale hydraulic fracture as it grows. Recently, a study aimed at assessing the potential for using multiple hydraulic fractures to pre-condition a rock mass to increase caveability and reduce fragment size provided just such an opportunity. The details of the site and the overall study results remain confidential as a condition of using the data, but the results of the stress change monitoring are of interest from a scientific perspective unrelated to the outcomes of the caveability study. This paper describes the results of three dimensional stress change monitoring about two full size hydraulic fractures induced in a crystalline rock mass. An initial hydraulic fracture was made using water as the injection fluid. This fracture was later reopened using cross-linked gel. 2. STRESS CHANGE MONITORING The stress change in the rock near the hydraulic fractures was measured using ANZI cells. These instruments are described in detail in Mills [1]. The ANZI cell is a soft, inflatable stress measurement instrument that measures the strain changes on the wall of a borehole induced by stress changes in the rock mass. In each instrument, eighteen electrical resistance strain gauges of various orientations are pressure bonded directly onto the borehole wall. A multiple linear regression analysis of the strain changes measured by the instrument allows the complete three dimensional stress changes to be calculated. A high level of redundancy (only six of the eighteen strain readings are required to determine the full stress change tensor) provides a strong measure of the confidence that can be placed in the integrity of the result.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper Number: ARMA-04-516

... stress measurement

**machine****learning**Artificial Intelligence classification primitive field stress stillwater complex excavation deformation field stress ABSTRACT: Primitive field stresses are the stresses that prevail today in the rock as a result of the original geologic formative...
Abstract

ABSTRACT: Primitive field stresses are the stresses that prevail today in the rock as a result of the original geologic formative processes including one or more tectonic episodes. It is commonly known in geological engineering that excavations change the primitive field stresses in close proximity. It is also known that the stresses so raised around excavations are a function of the magnitudes of the primitive field stresses and, if in excess of the strength of the rock, give rise to fractures in the rock. The degree of fracturing is directly related to the magnitude of the primitive field stresses. It represents the surface area of newly created fractures and can be assessed by the depth to which fractures are induced into the wall rock and the extent to which they occur around and along the excavation. Stillwater Mining Company developed a system for classifying the degree of observable excavation induced fracturing and for statistically relating such fracturing to the primitive field stresses in terms of the strength of the rock and the stress raising effect of the excavation. The observed degree of fracturing is empirically related in the literature to the proportion of the maximum stress to strength of the wall rock around an excavation. This proportion can be plotted against a similar proportion calculated from the geometry of the excavation and the rock parameters for a number of excavations. The ratio of the primitive field stresses to one another is one of the parameters on which the calculation is based. The value for the ratio of the primitive field stresses at which the average empirical and calculated wall stress to strength proportions are equal, is an estimate of the actual ratio of the primitive field stresses provided the associated correlation is acceptable. The system has been applied on the mine to three data sets and has resulted in estimates of the primitive field stress ratio that compare well with actual measurement and simulation of the incision of the Stillwater Valley in primitive landscape. The system is refined by ongoing application and extension of actual stress measurements across the Mine. 1. INTRODUCTION The Stillwater Mine, the largest primary producer of palladium and platinum in the Western Hemisphere, is situated in the Stillwater Igneous Complex in south central Montana, U.S.A. The Stillwater Complex is a 42 km long, layered, approximately 2.7 b.y. old lopolith that hosts a stratiform palladium and platinum rich layer known as the J-M Reef. Stillwater Mining Company has produced palladium and platinum from the J-M reef since 1987. Production is primarily accomplished by variations on overhand and underhand horizontal cut-and-fill methods with long-hole stoping practices playing an important role. Studies concerning the degree of mining induced fracturing began in earnest in early 2001. Such fracturing of the rock mass around excavations plays a crucial role in ground control methodologies. Determining the level of mining induced fracturing provides a tool to estimate the field stress state, which, in turn, can provide more reliable input parameters for design purposes.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 4th North American Rock Mechanics Symposium, July 31–August 3, 2000

Paper Number: ARMA-2000-0049

... project. It was found that the IKC tool, when mounted on a properly sized hydraulic breaker and operated by experienced people, can be production rate competitive with explosives for trenching, tunneling, vein mining

**machine****learning**reservoir geomechanics US government production control...
Abstract

ABSTRACT Viability of the impact-kerf-cutting (IKC) concept Impact rate of a tool and its effect on rock penetration within a velocity range of 10 to 20 ft/sec. Use of Schmidt Hammer for estimating unconfined compressive strength (UCS). The reliability of using UCS for production rate predictions. Production rate estimating procedures. ABSTRACT: This paper defines the principal aspects of primary and secondary rock fracture by impact. It discusses mechanical rock fragmentation and cutting tool design. And it investigates those physical rock properties that can be used to predict excavation rate performance. Items that are referred to for specific discussion are: This paper covers laboratory and field testing pertaining to the development of an impact slot-cutting/ridge breaking tool for excavating rock. This tool was developed for use on a hydraulic breaker with the intent of reducing the use of explosives in short or irregular shaped excavations. Laboratory drop testing was conducted at the Mining Equipment Research Facility (METF) of the US Bureau of Mines (Now NIOSH). Full-scale field tests were then conducted at two rock quarries and one foundation excavation project. It was found that the IKC tool, when mounted on a properly sized hydraulic breaker and operated by experienced people, can be production rate competitive with explosives for trenching, tunneling, vein mining

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 4th North American Rock Mechanics Symposium, July 31–August 3, 2000

Paper Number: ARMA-2000-0659

... problem. This paper looks at the issues surrounding variable weighting and standardization, and proposes algorithms for addressing the difficulties associated with it in fuzzy K -means cluster analysis. noise variable cooper 1988

**machine****learning**selection variable selection Artificial...
Abstract

ABSTRACT ABSTRACT: In the analysis of discontinuity data, it is often not known initially which variables are most useful in distinguishing between the different fracture groupings or sets. Cluster analysis tools are particularly vulnerable to the presence of variables that are essentially noise variables, because such variables can mask the true cluster structure of the data leading to poor cluster recovery. It is therefore essential to reduce the influence of unimportant variables by assigning importance weights to all variables involved in a cluster analysis. One other associated problem in cluster analysis is that the different measurement scales of different variables can cause some variables to dominate a cluster analysis solely because of the magnitudes of their readings and not because of importance. Variable standardization is the technique commonly used to solve this problem. This paper looks at the issues surrounding variable weighting and standardization, and proposes algorithms for addressing the difficulties associated with it in fuzzy K -means cluster analysis.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 4th North American Rock Mechanics Symposium, July 31–August 3, 2000

Paper Number: ARMA-2000-0667

... orientation roughness Artificial Intelligence Pattern Recognition Upstream Oil & Gas Euclidean Distance

**machine****learning**discontinuity data analysis quantization method bias index borehole algorithm cluster analysis discontinuity stereonet minimum value inclination angle...
Abstract

ABSTRACT ABSTRACT: This paper presents ongoing research in improvement of analytical tools for the characterization of rock mass structure from oriented borehole discontinuity data. Using automated multivariate cluster analysis and the "three dimensional stereo net" provides for fast and objective characterization of oriented core discontinuity data. Improvements to the CYL software include three new methods of cluster analysis, incorporation of roughness as well as orientation and spacing as variables, additional visualization modes, and an automated method to split the data set into different Geotechnical Mapping Units (GMU). In addition the program facilitates the selection of optimum drilling angle for boreholes.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 4th North American Rock Mechanics Symposium, July 31–August 3, 2000

Paper Number: ARMA-2000-1195

... estimation of parameters of the constitutive models through laboratory testing, it seems possible to obtain good agreements between the experimental and distinct element numerical results. Artificial Intelligence

**machine****learning**Upstream Oil & Gas mass strength criterion jointed model...
Abstract

ABSTRACT : To simulate soft rocks, (a) a mixture of plaster of Paris, sand and water was used as a model material. Thin galvanized sheets were used to create joints in blocks made out of the model material. To investigate the failure modes and strength, jointed model material blocks of size 30 x 12.5 x 8.6cm having different joint geometry configurations were subjected to uniaxial compressive loading. Results indicated three failure modes: (a) tensile failure through intact material, (b) a combined shear and tensile failure through joints, and (c) a mixed failure of the above two modes depending on the joint geometry. The fracture tensor of a jointed mass has the capability of integrating the effects of number of fracture sets, fracture density, and distributions for size and orientation of these fracture sets. The fracture tensor component in a certain direction provides the directional effect of the fracture tensor. Results obtained from the experiments were used to develop a strong non-linear relation between the fracture tensor component and the jointed model mass strength. Some of the laboratory experiments conducted on jointed model material blocks were simulated numerically using the distinct element method. Obtaining good agreements between the experimental and numerical results through the distinct element method was found to be not a trivial exercise. However, with careful selection of suitable material constitutive models for intact material and model joints and accurate estimation of parameters of the constitutive models through laboratory testing, it seems possible to obtain good agreements between the experimental and distinct element numerical results.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 2nd North American Rock Mechanics Symposium, June 19–21, 1996

Paper Number: ARMA-96-1185

... fracture directions, so they are expressing the regional geological history N020ºE, N040ºE(1)¸ N110-115ºE (2), N145-150ºE (3) and N060-080ºE (4), Figure 1. hydraulic fracturing orientation correlation matrix

**machine****learning**Upstream Oil & Gas eccentricity characteristic fracture network...
Abstract

ABSTRACT: Nowadays, the hydromechanical stability analysis and design of various workings in jointed rock masses are numerically performed by computing some more or less stochastic models. It is an increasing evidence that modeling/or simulating a natural fracture network needs to be as representative of the rock mass structure as possible. Not only average structural, geometrical and geostatistical characteristics have to be taken into account but also their variability. We propose an analysis of a fracture network, using size and shape of blocks outlined by a fracture pattern. Five size parameters and six shape factors of blocks outlined by quite vertical fractures on horizontal benches are computed ; their values were derived from 2095 traces resulting in 401 blocks. The statistical and multivariate analysis of these parameters give rise to an efficient classification of the blocks with respect to the chronology of the rock mass geological fracturation. INTRODUCTION The hydromechanical stability analysis and design of various working in jointed rock masses, such as rock slope, underground openings or dam foundations require the knowledge of the fracture pattern of the rock mass. It is a common purpose, analyzing ground water flow, to study a fracture network by the mean of structural and geometrical characteristics of the rock mass discontinuities (orientation, position, length, aperture, densities, spacing between joints, etc.). But, from a mechanical point of view, it is also possible and interesting to consider a jointed rock mass as a set of space filling blocks looking like polyhedra ; the faces, edges and vetices of the polyhedra being respectively distinct individual sub-planar fractures and intersection of two or three of them. So a rock mass is considered as a 3D mosaic of blocks, fractures being faces. In this paper we present an analysis of a fracture network of a rock mass located in the quarries of Comblanchien (Côte d'Or, France) using the size and shape of blocks outlined by the fracture pattern. As fractures are quite vertical, polyhedra become prisms and joint geometry is characterized by analyzing the size and shape of the two dimensional polygons out- lined on two partially superimposed horizontal benches -vertically 2.4 m apart-(Figure 1). 2 GEOLOGICAL DATA ACQUISITION Data acquisition (orientation and position) has consisted in a systematic manual in situ survey of the fracture traces. Co-ordinates of a set of points distributed on the fracture traces were measured with reference to an arbitrary orthogonal system (Gervais, 1993). Then 564 and 1531 traces were surveyed on 912 and 2075 m² area surfaces with densities LA (length per unit area) respectively equal to 0.92 and 1.06 m/m² (Gervais et al, 1993). Blocks are defined as closed domains outlined by fracture traces; if any ending part of a trace exists in such a domain, it is canceled. Finally we have 126 and 275 blocks (Figure 1). Principal orientation of polygon edges are parallel to the main fracture directions, so they are expressing the regional geological history N020ºE, N040ºE(1)¸ N110-115ºE (2), N145-150ºE (3) and N060-080ºE (4), Figure 1.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 2nd North American Rock Mechanics Symposium, June 19–21, 1996

Paper Number: ARMA-96-1567

... reference data of Cecil and others which were applied by Lee and Sterling, and Moon and Lee (1993). The inferred results, learning speed and ability were compared with those others. neural network deviation rock property algorithm

**machine****learning**node Lee output layer determination Sterling...
Abstract

ABSTRACT : Artificial neural networks were used to infer missing data when given geological data were insufficient. But the learning of the traditional networks were too slow for practical use. Authors adopted Fahlman's quickprop program to accelerated artificial neural network, to improve learning ability and speed up the learning time. Some test result showed learning time of new networks was reduced from tens of hours to a few minutes, while the output pattern was almost the same with other studies when the program ran on IBM compatible PC486 DX 66 MHz machine. Cecil's database was used for learning pattern. All the test run showed sufficient accuracy. 1. INTRODUCTION Accurate and resonable rock mass classification is needed to design and construct underground rock structures safely and economically. Rock classification methods are empirical ones to classify the rock mass and determine the support amounts by the ratings. Many items are needed like RQD, weathering condition of discontinuities, to classify the rock mass properly. Geological data sets can be usually insufficient in early stage of design because there are many limitations to reach underground site or uncertainty of the property itself. Traditional classification algorithms cannot work when one or more data items are omitted. Even all the data item is given, there can be a confliction between the data. In this case engineer must do some guesswork and inevitably be biased. Artificial intelligence like fuzzy system, expert system or artificial neural networks can be used to diminish the subjective errors. Neural networks are simplified models of neurons. They are composed of nodes in input, hidden and output layers and their connections. They learn facts from the given database through activation function and by controlling connection weights to make generalized inner information. They infer wanted data from new data set which they didn't learn or omitted or from distorted data set which they learned. There are more than 50 neural network models and many of them adopted error back propagation (EBP) learning algorithm with the generalized delta rule by Rumelhart et al.(1986) Some neural networks have been introduced in rock engineering field. Among them are networks to predict elastic compressibility of sandstone specimens (Zahng et al, 1991) and to infer failure mode of rock mass (Lee and Sterling, 1992). Conventional artificial neural networks require exceedingly long learning time. This disadvantage lead to lack of efficiency. They are too slow for practical use. In this study, authors take up new algorithms of accelerated artificial neural network, to improve learning speed and ability of artificial intelligent program. Adopted algorithm is Fahlman's quickprop program (Fahlman, 1989). It is modified and made to a module of ROMES, Rock Mechanics Expert System (Yang et al., 1995). New program was tested by well known XOR and character recognition problems to verify the validity. It was applied to the reference data of Cecil and others which were applied by Lee and Sterling, and Moon and Lee (1993). The inferred results, learning speed and ability were compared with those others.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 2nd North American Rock Mechanics Symposium, June 19–21, 1996

Paper Number: ARMA-96-1809

... & Gas Dershowitz

**machine****learning**fracture characterization mean pole Artificial Intelligence orientation iteration probability density function stochastic approach probability hydraulic fracturing Example application verification case Fisher algorithm orientation distribution geier...
Abstract

ABSTRACT : This paper presents an algorithm for definition of fracture sets based on a probabilistic, geological approach to fracturing. The algorithm asserts that fracture sets should be defined as groups of fractures with statistically homogeneous properties. There properties include orientation, but also emphasize geological information such as termination modes, displacements, striation, and mineralization. The sets are defined by assigning fractures to sets based on how similar all the characteristics of the fractures are to those of the other fractures in the set. The probability that a fracture is a member of a set is used to assign the fractures to sets. The properties of the sets are then calculated based on the statistics of the fractures assigned to the sets. Verification of the algorithm, and an example application are provided. 1. INTRODUCTION Rock mechanics analyses require definition of fracture sets based on geometric, geological, mechanics, and hydrological fracture properties. However, the most common technique for fracture set definition, stereonet contouring or cluster analysis (e.g., Shanley and Mahtab, 1974), relies exclusively on fracture orientations (see e.g., Fisher et al, 1987). In addition, contouring based algorithms assume that fractures are always assigned to the closest set pole, effectively truncating orientation distributions in areas of overlap. This can lead to biased estimates of variance. This paper presents a stochastic algorithm for fracture set definition, which has been successfully verified even for fractures sets with significant overlaps in orientation. This approach recognizes that every fracture identified in the field has a finite probability of membership in any given fracture set. Thus, the boundaries between fracture sets are probabilistic rather than rigid. 2. ALGORITHM The algorithm for fracture set definition is based on iterative, stochastic reassignment of fractures to sets, using updated estimates of a priori distributions for the properties of interest. At the outset, the a priori distributions of the properties of interest for each fracture set must be defined. For example, set one might be defined by a near horizontal orientation, with 0% termination at intersections, while set two might be defined by fractures with 100% termination at intersections. The algorithm must then evaluate every fracture and determine the probability that it is a member of either of the two sets. Fractures are assigned to sets based on their similarity to fractures already assigned to that set. This is achieved by assigning fractures based on their probability of being members of that set. This is referred to below as the "component membership probability, PJ k." The probability pj k denotes the component membership probability for fracture property j with respect to the a priori k-the fracture set. The algorithm begins by requiring weighting factors, Wj, for each of the HJ fracture characteristics of concern, and initial guesses for the properties of the each set k. For each fracture i, the probability that the fracture is a member of set k is calculated using the value of the probability density function for that property and set, taken at the value of that fracture.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 1st North American Rock Mechanics Symposium, June 1–3, 1994

Paper Number: ARMA-1994-0319

... on available project information. Artificial Intelligence

**machine****learning**criteria classification information construction method Upstream Oil & Gas Simulation database penetration rate utilization subsystem random number System Configuration equation excavation system...
Abstract

ABSTRACT ABSTRACT: This paper presents the design and development of a package, based on Monte Carlo simulation techniques, that can model the operation of mechanized underground excavation systems. The primary use of the package is to estimate time and cost required to complete a given tunnel, in addition to predicting equipment utilization. A data base containing performance records from more than 450 tunnel boring machine (TBM) projects is used to derive a ground classification system, predictive equations for various performance parameters, and probability density functions of various components of mechanized underground excavation systems. Analysis of various possible excavation system configurations is achieved by using rate of failure and delay information for each comprising subsystem. The estimation of time to complete a given tunneling project can be done in different ways depending on available project information.

Proceedings Papers

Publisher: American Rock Mechanics Association

Paper presented at the 1st North American Rock Mechanics Symposium, June 1–3, 1994

Paper Number: ARMA-1994-0793

... fractal. An example of the inverse problem of determining the fractal nature of unknown fault traces from seismicity is given. Artificial Intelligence Pattern Recognition hydraulic fracturing Barton point pattern fault trace experiment

**machine****learning**complex reservoir dimension fracture...
Abstract

ABSTRACT ABSTRACT: Fracture traces and seismic event location data were examined and numerical experiments performed to test whether scale-invariant fracture trace patterns can be represented by point patterns generated at discrete trace sites. Point pattern data censored in different ways were examined with respect to the effects on fractality. A sparse fractal point data set remains fractal, but the scale range of fractality decreases as sparseness increases. Thus, a seismic event location point pattern may not adequately represent the full scale over which the parent faults are fractal. An example of the inverse problem of determining the fractal nature of unknown fault traces from seismicity is given.