Upper Stillwater Dam is to be a roller-compacted concrete gravity structure, founded on nearly horizontally bedded sandstone and argillite rock. An overview of the testing and analyses used to evaluate the adequacy of the foundation relative to deformation, seepage, and stability is presented.
Upper Stillwater Dam will be the Bureau of Reclamation's first roller-compacted concrete gravity dam. It is located on the south flank of the Uinta Mountains in northern Utah. The maximum height of the dam is 82 m, and the crest length is 812 m at elevation 2492 m. An ungated overflow spillway will be constructed near the center of the dam. Water is diverted to Stillwater Tunnel or regulated to Rock Creek through a single intake structure. The general features of the dam are shown in figure 1. The dam will be founded on interbedded sandstone and argillite of the Precambrian Uinta Mountain group. The bedding structure is nearly horizontal at the site. The bedrock has been subdivided into five rock units for mapping purposes as follows: (1) an upper sandstone unit near the top of both abutments, (2) a middle sandstone unit with numerous interbeds of argillite and siltstone, (3) a thick argillite designated unit M extending to near the base of both abutments, (4) a lower sandstone unit which forms most of the foundation, and (5) a small but continuous argillite interbed designated unit L within the lower sandstone unit. A construction contract was per- formed to strip surficial materials, exposing seven minor faults designated F-1 through F-4 and F-7 through F-9. They are nearly vertical, cross the foundation from upstream to downstream, and generally consist of a gouge zone several millimeters wide within a fractured zone of rock about 1 m wide. Jointing, other than bedding joints, is limited to near vertical sets predominantly oriented parallel to the faults. joint spacings average between 0.5 and 3 m depending on the set and 1ocation. There is no distinct weathering profile, but most joints are filled with sand and organic materials to depths averaging 6 m. An erosional channel forms a bedrock low about [4 m deep near the right side of the foundation. This channel, termed the trough, should relieve any large horizontal stresses. Additional details of the geology are shown in figure 1.
Initial attempts to estimate in situ foundation deformation modulus were based on correlations with RMR (Rock Mass Rating, Bieniawski, 1978) and geophysical shear wave frequency. However, neither method was found to be totally acceptable for the rock at the Upper Still water damsite. In situ jacking tests were performed at the site utilizing the Goodman borehole jack. The data were reduced according to the method described by Goodman, Van, and Heuze (1968). Adjustment factors were also considered. However, the measured modulus values were low enough that corrections were considered unnecessary. The results from argillite were used directly because they were consistent and indicated a modulus parallel to the bedding consistently 1.5 times that perpendicular to the bedding. The sandstone results required further analysis.
A numerical method to derive approximate solutions for stress intensity factors for rectilinearly anisotropic bodies without resolving to complex variable approach is presented. The method is based on the assumption that the crack opening profile can be simulated by a simple curve such as an elliptic function and a relationship is found for the stress intensity factor in terms of near tip displacements. An example solution is derived for a transversely isotropic body and the results are verified with Westergaard's stress function approach. The proposed model has a great potential for the applications to fracture mechanics of anisotropic rock materials.
a Half crack length or major axis of a conic-section
all, el2 Elastic material constants
E Modulus of elasticity
G Shear modulus
KI Mode I stress intensity factor
_KI Normalized Mode I stress intensity factor
KII Mode II stress intensity factor
_KII Normalized Mode II stress intensity factor
K1 Mode I geometrical correction factor
K2 Mode II geometrical correction factor
k (3-4]>νννr, ¿ Polar coordinates
U1, v1 Displacements along and normal to the crack respectively
U, V Displacements in x and y directions respectively
x, y Cartesian coordinates
1, 2 x, y directions
a Angle of crack
sUniform tensile stress remote from a crack
sNormal tensile stress in the region of a crack
tShear stress in the region of a slanted crack
The field of fracture mechanics is increasingly being employed as an engineering approach to optimizing material fragmentation and engineering design. This requires knowledge of the stress intensity factor as a function of geometry, loading conditions, and material properties. Although many analytical solutions are available utilizing the classical theory of elasticity and highly sophisticated mathematical analysis, they are limited to idealized configurations and loading conditions. The area of mixed mode fracture in particular is inept in that the availability of solutions is, limited (Sih, 1973). Thus, it is important that different approaches are developed which are capable of determining accurate values of stress intensity factors (SIF) for a wide range of problems. This objective, however, is complicated by the fact that a singularity exists when finding the primary value of interest at the crack tip. An accurate and easily adaptable numerical technique is needed that will treat this singularity so that accuracy is maintained in the vicinity of the crack tip. The finite element technique is becoming well established as the best candidate in finding approximate stress intensity factors whenever an exact solution is not available. The method itself is conceptually simple, easily adaptable to high speed computers, and is applicable to a broad range of geometries, loading conditions, and anisotropic materials, with very good accuracy. The accuracy of the finite element stress analysis has been largely improved by using isoparametric finite elements and they are suitable for the applications to cracked bodies as well (Zienkiewicz, 1977). The use of special elements to model the singularity at the crack tip has largely improved the accuracy of the method in fracture mechanics applications (Tracey, 1974).
The resultant load vector is the representation of the forces applied to a longwall roof support element by caving strata into a single, quantifiable measure of support resistance. The relatively complex kinematics of the shield support prohibit a determination of support resistance simply by summation of leg forces. A method is being investigated by the Bureau of Mines to determine the resultant load vector by instrumenting supports with pressure transducers and strain gages to measure leg, canopy capsule, and lemniscate link forces. This concept has been laboratory tested in the Bureau's Mine Roof Simulator. Functional relationships among variables have been assessed, and confidence intervals have been established for prediction of the resultant load vector parameters.
RESULTANT LOAD VECTOR CONCEPT
Because of inadequate roof control and subsequent failures of several early longwall attempts in the United States, which utilized low-capacity European support equipment, there has been a tendency to increase support capacity with little regard to expected support loadings. Since the cost of a support is related to its capacity, use of excessively large supports represents an unnecessary capital investment and may cause unnecessary fracturing of the roof strata, thereby being detrimental to good roof control. Another major trend, which has occurred in American longwall mining practice during the past decade, has been the dominance of the shield support. The shield design, with its A-frame-type structure, offers the advantage of being able to resist horizontal loading with approximately 25 pct of the shield structure dedicated solely to this purpose (Peng, 1978). Although horizontal loading is a primary design consideration, the degree of horizontal loading and its association with geological conditions is relatively unknown. / Factors such as the uncertainty of horizontal and vertical loading demonstrate the need for additional in-mine support loading information to provide more effective design and utilization of longwall roof support systems. The Bureau is conducting research to provide a better understanding of support loading and caving phenomena by developing a technique to measure the resultant load vector on shield supports. The resultant load vector is the representation of the forces acting upon a support element by a single, quantifiable measure of support resistance. Being a vector, this measure possesses not only a magnitude, but also spatial parameters of location and direction as depicted in Figure 1. Reference will be made to three resultant load vector parameters:
Potential applications for boom tunnelling machines requires reliable assessment of in-situ performance. The majority of existing classification schemes concentrate on the prediction of excavation rate, neglecting the important influence of tool consumption rates. A statistical approach is taken to the analysis of the performance data, which was obtained first-hand through in-situ monitoring, and laboratory testing of representative rock samples. A statistical discrimination technique is used to produce a scheme for the prediction of performance groups, each of which covers a limited range of cutting rate and/or tool consumption rate.
Boom-type tunnelling machines have been in use in the UK for nearly 20 years. Throughout that time, assessment of the performance of machines prior to installation has always been demanded by the client. Research in the University of Newcastle upon Tyne and other establishments has proven a number of existing and novel methods of assessing the rock materials properties pertinent to excavation which, together with in-situ monitoring of machines, led to procedures for predicting machine performance. However, these procedures were only applicable over the very small range of comparatively light- weight machines available, and for single-strata, massive face conditions. Recent increased interest worldwide in the use of boom-type tunnelling machines, for excavation in Public Utilities Tunnelling, Mine Development, Underground Caverns for materials storage, and even Surface Mineral Production Operations, is demanding increasingly confident assessments for the in-situ performance of these machines prior to installation. The increase in the complexity in assessing performance for the wide range of strata now being excavated is further compounded by the range of machines now available. Thus a number of additional factors need to be considered in the equation for machine performance. Table 1 indicates the complex nature of the problem, which was discussed in more detail by Fowell R.J. and Johnson S.T. (1982). The object of this paper is to present a rational approach to machine performance prediction, introducing a practical and flexible scheme for performance prediction, using an approach which is closer to the considered qualitative approach utilised by the experienced engineer. The present scheme aims to effectively quantify this assessment process. In order to attempt to produce such a scheme, data is needed. The data used in this work was obtained from in-situ monitoring of a number of machines, both trackmounted machines, and where the boom was mounted within a shield. The former configuration is preferred when arch-shaped girder supports are used, typical of mine development drives. The latter for smooth profile circular tunnels for sewage schemes, etc., where an immediate sectional concrete support/ lining is desired. Performance is predicted in terms of a machine cutting rate (m3/h), which is independent of the operating system, the type of contractor, and the management of the workforce, and thus is dependent on the machine specification, the rock material and rock mass properties. Tool consumption also is effectively dependent on the same parameters, although results taken are presented as shift averages.
APPROACHES TO MACHINE PERFORMANCE PREDICTION
A brief examination of some of the more popular methods in use for the prediction of machine performance resulted in a broad classification.
Amadei, B. (University of Colorado, Dept. of Civil Engineering) | Janoo, V. (University of Colorado, Dept. of Civil Engineering) | Robison, M. (University of Colorado, Dept. of Civil Engineering) | Kuberan, R. (Central Soil and Materials Research Station)
Laboratory studies on the deformability and strength of intact rocks generally involve the uniaxial compression test, the standard triaxial compression test and direct or indirect tension tests. If sl, s2, s3 are principal stresses (s1 > s2 > s3) with compression taken as positive and tension taken as negative, the applied stress field for each one of these tests can be represented as shown in Fig. 1. However, for most in-situ conditions, such as those that exist at any point around an underground excavation, the stress field is truly three-dimensional or multiaxial (s1 ¿s2 ¿s3) and none of the tests shown in Fig. I can be used to predict the mechanical behavior of the rock. There are special cases for which one of the three principal stresses vanishes. Biaxial loading of rocks in-situ takes place, for instance, when rocks deform and fail in plane stress conditions. This can occur in a free unloaded surface of a rock structure such as in the wall of an underground excavation or a rock slope. Parallel to the free surface, the stress field components can be both compressive (s1>s2 >0), both tensile (0 >s2>s3)), or mixed, one component being compressive and the other one tensile (sl >0>s3 ). The purpose of this paper is to present the results of 25 true biaxial tests that have been conducted on cubical specimens of Indiana Limestone using the multiaxial cell available at the University of Colorado at Boulder. The tests include uniaxial compression and tension, biaxial compression and combined biaxial compression-tension. Flexible fluid cushions and brush bearing platens were used for compressive and tensile loading respectively in order to minimize loading end effects on the test specimens. The paper begins with a review of previous investigations on the multiaxial behavior of rock. Then, test results are presented in terms of biaxial strength characteristics of the limestone. Finally, these characteristics are compared to those derived from conventional tests such as those shown in Fig. 1.
The behavior of rocks under unequal stresses (biaxial or triaxial) has been investigated by testing (i) hollow cylinders of rock under different boundary loading conditions such as axial load combined with inner and/or outer confining pressures or axial load combined with torsion and confining pressure, and, (ii) shaped specimens ("dog bone" specimens). A review of the test procedures and a summary of test results can be found in Jaeger and Cook (1976). These tests were used to study the behavior of rocks in uniaxial and biaxial compression and in combined compression-tension but suffered several major disadvantages. The first one is that the stress distributions within hollow cylinder specimens must be assessed from the theory of linear elasticity and are non-uniform unless thin- walled hollow cylinders are used. The non-homogeneous stress distribution leads to the second disadvantage which is the difficulty in defining exactly the stress condition at failure. Problems associated with testing shaped specimens concern specimen preparation, stress concentration at the specimen ends and loading conditions.
Because of the hazards associated with flammable methane gas and coal dust, the shooting of mudcaps (adobes) or other unconfined explosive charges in underground bituminous coal mines is prohibited; all explosives must be fired in stemmed boreholes. However, there are situations where it would be advantageous from a safety standpoint to fire open shots. This would be in the areas of dislodging loose roof slabs, overhangs, rock-fall leveling, slab or boulder breaking, crib removal, and crevice shooting. The Bureau of Mines, U.S. Department of the Interior, has been developing an explosive charge that is non-incendive; that is, detonation of the charge will not ignite a flammable atmosphere. This explosive charge could be applied to the blasting conditions mentioned above. As of this writing, the explosive charge has not yet been certified as "permissible" for use in flammable atmospheres, but the necessary protocol is being finalized with the Mine Safety and Health Administration (MSHA), U.S. Department of Labor. The explosive charge described in this paper is merely a prototype sheathed rock-breaker charge (sheathed permissible explosive rockbreaker charge) demonstrating the feasibility of such a device. Industry may modify the design and composition of the charge in producing a commercial product so long as it is still safe, as defined by a testing schedule being developed by the Bureau of Mines and MSHA.
EXPERIMENTAL PROCEDURES AND RESULTS
Research conducted by the Bureau of Mines has shown that a properly designed, prepackaged explosive charge could be safely fired openly in a flammable atmosphere if the explosive were covered with a layer of sodium chloride (NaC1) that would be dispersed as a fine cloud upon firing. For an explosive charge of 650 grams, equivalent to two 3.2- by 40-cm cartridges of permissible water gel explosive, a 1.3-cm layer of NaC1 proved adequate (Mainiero and Hay, 1982). Included in this research was an investigation into the proper shape for an explosive charge that would effectively break stone. Charges of various shapes with lined and unlined cavities were tested with the result that these were inefficient rock breakers. The most effective charge proved to be one in the form of a short cylinder 17.8 cm in diameter and 2.2 cm high (not including the NaC1 layer). This shape spread the explosive over a large surface area of the rock, yet provided a thickness adequate to provide for efficient detonation. Based on the research described above, the charge illustrated in figures 1 and 2 was developed. The explosive charge, packaged in a polystyrene container, is covered with a 1.3-cm layer of damp NaC1 (88 pct Nacl-12 pct water), which in turn is encased in a housing consisting of latex rubber reinforced with cheesecloth. This configuration provides a charge package that is soft enough to conform to the irregular surface of a rock yet strong enough to withstand rough handling.
Figure 1. Diagram of sheathed explosive rock-breaker charge. (available in full paper)
Figure 2. Exploded view of sheathed explosive rock-breaker charge (available in full paper)
The incendive characteristics of the new sheathed rock-breaker charge in flammable gassy atmospheres were evaluated by firing charges against concrete slabs in a cylindrical steel gallery 1.8 m in diameter, 3 m long.
The G-Tunnel heated block experiment is being conducted on the Nevada Test Site (NTS) as part of the Nevada Nuclear Waste Storage Investigations project (NNWSI). The purpose of the ambient temperature testing phase is to evaluate rock-mass mechanical properties of a block (~8 m3) under biaxial stress changes up to 7.5 MPa above an initialization in situ value of 3.1 MPa. Results indicate that the modulus of deformation ranges from 9.7 to 17.0 GPa and Poisson's ratio ranges from 0.21 to 0.33. In general, the higher values of the modulus and Poisson's ratio were influenced by fracture propagations parallel to the compressive stress field. Other measurements indicated that cross-hole compression (p) wave velocities and single fracture permeability values were relatively insensitive to stress changes above the in situ value.
Volcanic tuffs are being considered by the Department of Energy (DOE) for the possible disposal of commercial high level radioactive wastes. The Nevada Nuclear Waste Storage Investigations (NNWSI) project was established in 1977 to evaluate such disposal in geologic formations on or adjacent to the Nevada Test Site (NTS). Sandia National Laboratories (SNL), as one of the participants of the NNWSI project, is responsible for the rock mechanics program to support the design and evaluation of a repository in tuff. The data from this heated block experiment provide key input for definition of the rock-mass mechanical properties to be used in the conceptual design. The rock mechanics field program has begun in G-Tunnel, where tuffs similar to those at the potential candidate site are found; later experiments are planned as part of the exploratory shaft investigations at Yucca Mountain, the potential site. Science Applications, Inc. (SAI) has been under contract to SNL to provide assistance in the design, installation, operation, and evaluation of this experiment. They have been responsible for the preparation and installation of the instrumentation and the control systems as part of this effort. Related details will be presented separately from this paper. The purpose of this paper is to present the results from rock-mass mechanical measurements taken prior to the application of heat to the block. The heated block is a multi-faceted experiment designed to evaluate rock-mass thermal (thermal conductivity), mechanical (modulus of deformation, Poisson's ratio, joint shear and normal properties), thermomechanical (coefficient of thermal expansion), and hydrological (fracture permeability, moisture content, pore pressure) responses on a relatively large scale (~8 m3) for conditions where stress and temperature boundary conditions can be reasonably well controlled. The heated block is located in the floor of an alcove driven from the Rock Mechanics Drift in the G-Tunnel Under- ground Facility. The alcove is in welded tuff and was mined using standard drill and blast techniques. The block was formed by cutting four 3-m-deep orthogonal slots in the floor of the alcove to form the periphery of a prism of jointed-welded tuff 2 m square. The slots were cut with an overlapping line drilling method using percussion drilling techniques. Flatjacks were grouted into the slots in configurations shown in Figure 1.
Computer assisted finite element model analysis, originally developed to study stress distributions in aircraft, has been modified into a versatile mining program, that can be used to assist the engineer with predictions of mine subsidence. International Exploration, Inc. has developed an in-house finite element model program which incorporates a multitude of factors influencing mine subsidence such as mine design, physical rock properties, structure, faults, joints and topography. Finite element analysis of a mine model can mathematically determine the magnitude and direction of the principal stresses within each of the elements forming the model. By comparing the calculated in-situ stresses to the actual rock strengths, high risk regions within a mine can be identified. These potentially unstable regions of the mine can be allowed to fail within the model and the resultant stress redistribution and element movement calculated. Thus, the effects of the mine collapse at the surface can be predicted for the purpose of determining the potential damage, if any, that could occur to surface structures from deep mine failure. Stabilization materials can also be mathematically inserted into the model to design and analyze the effectiveness of subsidence prevention programs. Two examples are presented of the finite element program being used' to predict mine subsidence at substantially different geologic and mining provinces in Pennsylvania.
Abandoned subsurface coal mines in Pennsylvania have posed a substantial problem to hundreds of property owners situated above them. The mines may collapse, with time, from various causes such as weathering, pillar robbery or mine flooding. The resulting subsurface collapse may cause significant surface movement of a magnitude and direction capable of damaging surface structures. As a result, mine stabilization projects are faced with determining which areas within hundreds of square miles warrant special consideration. Through the use of computer assisted finite element modeling, specific high risk areas can be delineated and recommendations suggested for developing strategies to prevent future damage to surface structures. Using finite element modeling, the mining and geologic conditions of a specific study site are mathematically recreated to calculate the stresses present within mine pillars and rooms. Over stressed areas within the model can then be caused to fail and the resulting magnitude and direction of surface movement and stress then determined. The mathematical model can be modified to consider the effects of pillar decay and robbery as well as introduce faults, secondary mining and stabilization materials. The ability of the program to calculate the stress conditions within a mine under the complex situations commonly found in Pennsylvania greatly exceeds the capabilities of empirical approaches.
FINITE ELEMENT MODELING
The finite element method is a general mathematical technique of structural analysis in which the body to be studied is subdivided into individual structural elements, in this case, triangles (Zienkiewicz, 1977) (Figure 1). The triangles are interconnected at .their corners or nodes. When forces are applied to the body, the nodes displace, and the triangles experience strain. The amount of displacement of each triangle depends on the level of the forces applied to the triangles and the material properties of each triangle.
This paper describes a number of model tests conducted in plexiglas models to investigate the phenomenon of fracture pressurization. The models were examined with high speed photography while being subjected to explosive loading. At the same time pressure transducers were used to record the pressure in the borehole as a function of time and also along the path of the propagating fracture to measure the pressure at various locations along the fracture as a function of time. Both propellants and explosives were used to charge the borehole. Air as well as fluid filled boreholes were needed to provide a variety of pressure rise rates. On some tests eddy current displacement gauges measured crack opening displacements as a function of time. As a final check high speed photographs taken during the event were used to visually ascertain the location of the fracture at any given time.
The speed with which fractures created by an explosive detonation are filled with high pressure gases as well as the magnitude of the pressure in the fractures are of great interest. To date no valid data has been presented that sheds light on this very complex event. In the case of fragmentation blasting it has been postulated that the amount of rock breakage that results is very much a function of successful pressurization of the fractures. The exact mechanism of fragmentation is unknown but the current theory is that in a matter of microseconds an intense fracture network is created in the near vicinity of the borehole. At some as yet undetermined time later (tens or hundreds of milliseconds) these fractures are filled with high pressure gases which continue to drive the fractures and jumble the resulting rock fragments. The proper combination of stress waves and gas pressures result in good fragmentation. Once this process is at a certain stage then proper blasting procedures call for a second hole or series of holes to be detonated. Before proper fragmentation blasting can be planned a complete knowledge of the pressurization process, how it is affected by pressure rise rate and at what time it occurs, must be determined. In other areas of blasting practice it is also important to understand the process of fracture pressurization. In oil and gas well stimulation with explosive and propellant charges it is desired to create multiple fractures which travel from the borehole wall and intersect natural fracture systems within the reservoir in order for .the trapped hydrocarbons to flow into the well bore so that they can be taken to the surface. In this application it has been demonstrated that if gases which are created by the explosion do not penetrate into the stress wave created fractures very little production is achieved. Although sophisticated computer codes exist for predicting well bore fracturing, S. L. McHugh, et al 1978, and rock fragmentation, T. G. Barbour et al 1980 and S. L. McHugh 1980, they have been ineffective since no physical model is available to predict the crack pressurization event.
Characterization of the rock mass in which deep underground structures are to be located is of critical importance to the feasibility and design of the structures. The process involves area1 studies, subsurface exploration and extensive testing to obtain comprehensive data on the rock units which will enclose the excavated structures. The Illinois Deep Hole Project provides an example of characterization of rock at great depth describing the types and variety of tests and studies which were utilized and the results obtained.
Rock characterization is a detailed description of the nature and quality of the rock integrated from data collected from many different sources. The character of the rock is related to the nature and behavior of the rock itself and of the rock mass as a whole with its system of discontinuities. Some characteristics can be explored by indirect methods from the ground surface; some require physical testing of samples of rock; others necessitate in-situ testing of the rock mass at depth. The studies described illustrate some of the investigations which can be used to characterize rock for large underground structures. The initial purpose of the Illinois Deep Hole Project was to select a site and perform subsurface investigations for an underground pumped storage project. The project was to include six to eight shafts up to eight meters in diameter, an underground power station, approximately 25x50x260 meters, and about 19 kilometers of tunnels, with diameters of about 25 to 30 meters, comprising the underground reservoir. To enclose the structures for this project a massive, competent, relatively impermeable rock unit at a depth of about 1500 meters was required. The only rock in northern Illinois which could meet these qualifications was Precambrian granite. For a project of such magnitude, a thorough characterization of the rock mass was essential, and the Illinois Deep Hole Project was unique in the comprehensiveness of the geologic and geophysical studies and considerable additional research performed by various members of the academic community. The studies performed in the characterization were divided into four major categories, 1) areal geophysical surveys, 2) core logging, 3) downhole testing, and 4) laboratory studies. The various tests and methods used and objectives of the tests are listed in Fig. 1.
AREAL GEOPHYSICAL SURVEYS
The project area was selected initially on the basis of fragmentary structural geologic data showing the Precambrian rising to the north toward the Wisconsin Arch and of surface gravity data indicating a negative gravity anomaly, often correlating in the midwest to a granitic high. Later regional surface gravity surveys confirmed this anomaly and outlined an oval-shaped granitic body approximately five kilometers thick (Aiken, et al., 1983). These data also correlated with later-released aeromagnetic data and heat-flow studies. The project is located in an area of low seismicity between the Plum River Fault, approximately 35 kilometers to the south, and an unnamed fault, about 24 kilometers to the north. The site is about 53 kilometers from the northwestern end of the major Sandwich Fault. A Vibroseis reflection survey, comprising 105 kilometers of line in a 80-square-kilometer area, was run to determine the configuration on top of the Precambrian and the thickness of the Mr. Simon, a thick water-bearing sandstone overlying the Precambrian.