The effect of frac-hit among the stimulated horizontal wells located in the northwest of the State of New Mexico are identified by addressing how to predict whether or not a planned well caused frac-hit for older wells nearby, and in case of the frac-hit occurrence, how to predict the degree of impact. The machine learning method is used to find the relationship between well parameters such as distance and age difference, and frac-hit occurrence and the degree of impact. Determining the probability of frac-hit occurrence is considered as a classification problem, and random forest method is used to predict the occurrence of the frac-hit. Predicting the impact of the frac-hit is considered as a regression problem, and two machine learning methods, gradient boosting and adaptive boosting (AdaBoost), are used to solve this problem. In the pool of data, the data are randomly assigned to train and test set for unbiased machine learning.
The data of the training set are put into the random forest classifier to find whether the distance, age, age difference, and bearing have any impact on the occurrence of the frac-hit. Among these four factors, the bearing has the most significant impact, which means that the weight of bearing in classification process is higher than the other parameters, followed by the distance as the second important factor. Applying the trained random forest classifier on the test set data gives 78% correct outcomes compared to the actual frac-hit data in the test set.
Considering the change of oil production due to frac-hit as the indicator to measure the degree of impact in gradient boosting and AdaBoost algorithm shows that the bearing between wells is not an influential parameter in the regression problem compared to the classification problem. In other words, if the well has already experienced the frac-hit, the importance of bearing decreases, and the distance, age difference, and age of the wells become more prominent factors. The analysis shows that the average error between the actual data and the predicted results by gradient boosting and AdaBoost is about 40%.
The results of this paper can be used by the hydraulic fracturing operators to pre-determine the frac-hit probability and its impact on existing offset wells. It can also help to refine well design strategies to minimize the risk of potential well interferences.
Seright, Randall S. (New Mexico Institute of Mining and Technology) | Wang, Dongmei (University of North Dakota) | Lerner, Nolan (Cona Resources Limited) | Nguyen, Ahn (Cona Resources Limited) | Sabid, Jason (Cona Resources Limited) | Tochor, Ron (Cona Resources Limited)
This paper examines oil displacement as a function of polymer-solution viscosity during laboratory studies in support of a polymer flood in Canada’s Cactus Lake Reservoir. When displacing 1,610-cp crude oil from field cores (at 27°C and 1 ft/D), oil-recovery efficiency increased with polymer-solution viscosity up to 25 cp (7.3 seconds-1). No significant benefit was noted from injecting polymer solutions more viscous than 25 cp. Much of this paper explores why this result occurred. Floods in field cores examined relative permeability for different saturation histories, including native state, cleaned/water-saturated first, and cleaned/oil-saturated first. In addition to the field cores and crude oil, studies were performed using hydrophobic (oil-wet) polyethylene cores and refined oils with viscosities ranging from 2.9 to 1,000 cp. In field cores, relative permeability to water (krw) remained low, less than 0.03 for most corefloods. After extended polymer flooding to water saturations up to 0.865, krw values were less than 0.04 for six of seven corefloods. Relative permeability to oil remained reasonably high (greater than 0.05) for most of the flooding process. These observations help explain why 25-cp polymer solutions were effective in recovering 1,610-cp oil. The low relative permeability to water allowed a 25-cp polymer solution to provide a nearly favorable mobility ratio. At a given water saturation, krw values for 1,000-cp crude oil were approximately 10 times lower than for 1,000-cp refined oil. In contrast to results found for the Daqing polymer flood (Wang et al. 2000, 2011), no evidence was found in our application that high-molecular-weight (MW) hydrolyzed polyacrylamide (HPAM) solutions mobilized trapped residual oil. The results are discussed in light of ideas expressed in recent publications. The relevance of the results to field applications is also examined. Although 25-cp polymer solutions were effective in displacing oil during our corefloods, the choice of polymer viscosity for a field application must consider reservoir heterogeneity and the risk of channeling in a reservoir.
ABSTRACT: The interaction of lithology and different geological units (e.g., joints, beds, faults, and alteration) and their impacts on rock slope stability can impose significant challenges for a mining operation. Numerical modeling is a powerful tool for simulating fault response. This paper attempts to highlight the challenges in simulating faults in numerical modeling using FLAC3D software and ways to overcome these problems. A case history for a fault slip was used to assess the performance of present methods (weak zone, ubiquitous-joint, and interface) for fault modeling. Strength reduction method was used to evaluate the stability of the slope. A sensitivity analysis was conducted to investigate the impact of fault geometry and thickness on the overall stability of the slope. It is found that the fault responds differently using each method. The lowest FOS value were observed in weak zone method, whereas the ubiquitous-joint technic showed the highest FOS value. Furthermore, the mechanism of slope failure was slightly different using each simulation method. The reason for this is whether the fault is modeled, so tight that has no reaction to mining or alternatively, is moving in an uncontrolled manner. Therefore, it is important to calibrate the model for correct fault movements.
Determining slope stability is a critical part of the design stage of mining and civil engineering projects. This is especially important in large open-pit mines and dam construction. In an open pit mine, a slope failure not only results in a delay in production, also may cause fatality and equipment loss (Fellenius, 1936; Duncan & Goodman, 1968; Mahtab & Goodman, 1970; Stacey, 1970; Stacey, 1972; Stacey, 1973; Duncan & Wright, 1980; Hoek & Bray, 1981). Nowadays, with the growth of the global demand for minerals and advancements in technology, the depth of open-pit mines has increased significantly. This has imposed several design challenges in slope engineering and has exponentially increased the possibility of large slope failures (Kalkani, 1975; Stacey, 1996; Sjoberg & Norstrom, 2001; Wines & Lilly, 2002; Wyllie & Mah, 2004; Franz, 2009; Tutluoglu, et al., 2015)
The factor of safety (FOS) is a common approach for evaluating the stability of a slope. Bishop defined the FOS as the ratio of actual shear strength to the minimum shear strength required to maintain equilibrium, Fig. 1). (Bishop, 1955; Matsui & San, 1992; Fleurisson, 2012; Abderrahmane & Abdelmadjid, 2016). Hence, the FOS value of one (FOS=1) shows that the failure is imminent.
ABSTRACT: Dynamic uniaxial compressive strength of Pennsylvania blue sandstone was investigated using split Hopkinson pressure bar both physically and numerically. A hybrid finite-discrete element code called CA3 was employed to simulate the physical tests. The incident and transmitted bars were modeled using finite elements while the rock specimen was represented by a bonded particle discrete system. The incident stress pulse measured in the physical test was utilized as the input for the numerical simulation and was applied to the free end of the incident bar. Analysis of the numerical results suggests an underestimation of the dynamic rock strength; the effect of axial and circumferential inertia of the specimen didn’t manifest the strength value consistent with the physical observation. Therefore, a parameter called rock strength enhancement coefficient was introduced which increases the bond strength between the particles as a function of the relative velocity of particles at the contact points. A much better match between the physical and numerical results is observed if this coefficient is applied in the numerical simulation.
Most of the operations on the rock like materials from mining and road structures to dam foundations usually include the dynamic application of the load to the rock. Rocks are pressure sensitive and rate-dependent materials and show a drastically different behavior under dynamic loading. Since the dynamic loading of rocks is applied in a variety of loading rates, it is essential to study the dynamic strength parameters of the rocks and fracture properties over a wide range of loading rates.
There are three main methods for testing the rock materials under dynamic loading conditions which have been suggested by the International Society for Rock Mechanics . These methods include dynamic compression test, dynamic Brazilian test, and dynamic notched semi-circular bend (NSCB) test. All these tests are performed using the split Hopkinson pressure bar (Kolsky bar).
The split Hopkinson pressure bar (SHPB) was first designed for testing of ductile materials such as metals . In the last couple of decades, SHPB has been widely used for evaluating various parameters for brittle materials like rock, concrete, and ceramics [3, 4].
Along with the physical tests, many numerical methods have been utilized to investigate the characteristics of quasi-brittle materials such as rock and concrete in dynamic loading. Various constitutive models were used in ABAQUS to study different parameters which affect the dynamic strength of rock like materials in the FEM method [5, 6, 7]. As previous studies show, discrete element method (DEM) provides an alternative and reliable solution for discontinuum materials like rocks [5, 8, 9]. However, this method is computationally time-consuming when it comes to large scale domains.
Brattekås, B. (The National IOR Centre of Norway, University of Stavanger) | Ersland, G. (University of Bergen) | Seright, R. S. (New Mexico Institute of Mining and Technology)
Crosslinked polymers extrude through fractures during placement of many conformance improvement treatments, as well as during hydraulic fracturing. Dehydration of polymer gel during extrusion through fractures has often been observed, and was extensively investigated during the last decades. Injection of highly-viscous gel increases the pressure in a fracture, which promotes gel dehydration by solvent leakoff into the adjacent matrix. The present comprehension of gel behavior dictates that the rate of solvent leakoff will be controlled by the gel and fracture properties, and to a less extent impacted by the properties of an adjacent porous medium. However; several experimental results, presented in this work, indicate that solvent leakoff deviates from expected behavior when oil is present in the fracture-adjacent matrix. We investigated solvent leakoff from Cr(III)-Acetate-HPAM gels during extrusion through oil-saturated, fractured core plugs. The matrix properties were varied to evaluate the impact of pore size, permeability and heterogeneity on gel dehydration and solvent leakoff rate. A deviating leakoff behavior during gel propagation through fractured, oil-saturated core plugs was observed, associated with the formation of a capillary driven displacement front in the matrix. Magnetic Resonance Imaging (MRI) was used to image water leakoff in a fractured, oil-saturated carbonate core plug and verified the position and existence of a stable displacement front. The use of MRI also identified the presence of wormholes in the gel, during and after gel placement, which supports gel behavior similar to the previously proposed Seright filter-cake model. An explanation is offered for when the matrix impacts gel dehydration and supported by imaging. Our results show that the properties of a reservoir rock may impact gel dehydration; which, in turn, strongly impacts the depth of gel penetration into a fracture network, and the gel strength during chase floods.
Kumar, Abhash (National Energy Technology Laboratory) | Zorn, Erich (National Energy Technology Laboratory) | Hammack, Richard (National Energy Technology Laboratory) | Harbert, William (University of Pittsburgh) | Ampomah, William (New Mexico Institute of Mining and Technology) | Balch, Robert (New Mexico Institute of Mining and Technology) | Garcia, Leonard (New Mexico Institute of Mining and Technology)
A surface seismic network of five broadband seismometers was deployed in the vicinity of a CO2 injection well at an active enhanced oil recovery site in Farnsworth, Texas. We examined and characterized the data collected during the first three months of deployment. Data analysis identified 280 high-amplitude, regional events and a second set of 12 long period, low frequency events. The hypocenters of high-amplitude, regional events are distributed throughout central to western Oklahoma and are unrelated to CO2 injection in the Farnsworth field. The nearest cluster of events in western Oklahoma is >90 miles away from the injection well in Farnsworth. Long period low frequency seismic events observed in this study have emergent waveform characteristics that persist for 30-70 seconds. Spectral analyses of these events revealed a significant concentration of energy in the 0.8-3 Hz frequency range. A finite temporal moveout is observed in the arrival time of recorded long period events across the local seismic array in Farnsworth. These low frequency events have waveform characteristics significantly different from a short period earthquake and resemble long period, long duration (LPLD) events previously reported. The absence of time-correlative signal from the nearby stations of the Oklahoma seismic network and arrival time moveout recorded across the Farnsworth seismic network suggest a local source of slow slip deformation as the cause for long period events observed in this study.
Presentation Date: Wednesday, September 27, 2017
Start Time: 3:55 PM
Presentation Type: ORAL
ABSTRACT: Providing efficient support system which prevents failures and deformations is one of the most important issues during tunnel construction especially in weak grounds. When conventional support systems such as rock bolts, shotcrete, wire mesh, steel frames and lattice girders cannot provide sufficient support for tunnels, using pre-reinforcement systems becomes necessary in addition to main support systems. Pre-reinforcement of weak ground is done before the excavation or ahead of advancing the tunnel. This will provide a safe and effective operation. Pipe roofing umbrella arch pre-reinforcement method is one of the conventional pre-reinforcement systems that can be implemented in tunnels, caverns and other infrastructures construction. Detailed 3D numerical simulations are useful tools to obtain a better understanding of the performance of a pre-reinforcement system. In this paper, sectional excavation of tunnel No.10 of the Ghazvin-Rasht Railroad is simulated with FLAC3D code by using pipe roofing pre-reinforcement method, side bolts and pipes, and initial support system. The results of numerical simulations are analyzed by using tunnel support interaction charts. Considering all technical parameters, pipes with 4 in diameter, 15 cm spacing and 5 degree installation angle is selected as the most appropriate pipe roofing method for this tunnel.
Tunneling in soft or weak ground may encounter many operational and safety problems. Under these conditions extra attention and field investigations will be required. When rock mass cannot be self-supporting, or when conventional support systems such as rock bolt, shotcrete, wire mesh, steel frame or lattice girder cannot provide sufficient support for a tunnel, using secondary support systems or pre-reinforcing become necessary. Furthermore, in urban areas shallow tunnels are often constructed adjacent to existing structures such as buildings, streets and railways (Funatsu et al., 2008). The possibility of tunnel induced displacements and subsidence that can damage the existing structures is very high. In order to minimize effects of tunneling on existing structures, engineers should pay special attention to ground displacements. Therefore, using efficient and proper support system is necessary with considerations of the project requirements and constraints such as amounts of settlements and displacements. Often, it is required to use secondary support systems or pre-reinforcing in addition to primary support system in order to ensure a safe working place during the tunnel excavation. It is obvious that if effective support system is not installed at proper time and place, tunnel will collapse and may cause extensive damage, resulting financial and human loss in the project.
ABSTRACT: The bonded particle model is a powerful tool in studying the mechanical behavior of rock. The common practice in simulation of rock failure using this model is to allow brittle fracture of the contacts between particles or at most tensile softening at the contact points ignoring the shear softening of the material. To overcome this shortcoming, a plasticity model that allows both tensile and shear softening of the filling material at the contact points of the particles was implemented in the CA2 computer program. The model was calibrated to mimic the elastic behavior of the Pennsylvania blue sandstone. It is shown that for a more ductile material, there is less scatter of micro-cracking at the peak load. Furthermore, the ductility parameter appears to be a good tool in controlling the ratio of compressive to uniaxial tensile strength of rock. While the ductility of the filling at the contact points of the particles has a drastic effect on the macroscopic post-peak rock behavior in the direct tensile testing, its role in dictating the post-peak rock behavior in compression is negligible and needs further study. The combined effect of ductility and initial micro-cracking on rock strength characteristics was studied as well. The numerical results suggest that the ratio of Brazilian to direct tensile strength of the simulated material is affected by the initial micro-crack intensity; this ratio is around 1 for a material with no initial micro-cracks but it gradually increases as the initial micro-crack intensity is increased.
The Bonded Particle Model (BPM) has found a variety of applications in rock engineering and rock mechanics owing to its simple constituent analogy to circular discs in two dimensions and spheres in three dimensions and its ability to model complex rock and soil behavior using simple contact models for interaction of the particles (Potyondy and Cundall, 2004; Norouzi et al., 2013; Fakhimi and Hemami, 2015; Lisjak and Grasseli, 2014). While the BPM has been considered as a great tool for geotechnical modeling, it has its own shortcomings and drawbacks. One of the challenges in BPM modeling of rock is the problem with obtaining realistic values for friction angle and ratio of uniaxial compressive strength (UCS) to tensile strength; the friction angle and UCS values are normally underestimated by the BPM model using spherical or circular disks (Fakhimi, 2004; Wu and Xu, 2016).
Axial splitting is an important failure mechanism in rock engineering. This failure mechanism is particularly observed in the vicinity of free surfaces of rock structures. Different possibilities such as existence of defects, pores and micro-cracks in the rock specimen have been attributed as the cause of axial splitting in the literature. Based on some physical uniaxial compression tests on Pennsylvania blue sandstone, an alternative theory for axial splitting is proposed. In this theory, axial shear cracks are considered that are induced because of non-uniform deformation of the specimen ends. Dislocation along these axial shear cracks is the cause of rock lateral dilation which results in radial stresses. The radial stresses push out the cylindrical thin shells of the specimen. Consequently, radial cracks are developed which subsequently cause rock spalling in mode-I fracture. The proposed failure mechanism is discussed within the framework of a simplified theoretical model. The prediction of the model is then compared with the physical observations. In addition, a discrete element bonded particle system is utilized for rock simulation in uniaxial compression testing. The prediction of the numerical model is consistent with the physical observations.
Axial splitting is a complicated failure mechanism that is normally observed in uniaxial compressive testing of rock. Wawersik and Fairhurst  conducted some uniaxial compressive tests and defined local fracturing that is mostly parallel to the specimen axis and both local and macroscopic shear faulting as two possible failure modes. In the work of Horii and Nemat-Nasser , a pre-existing inclined crack in the rock specimen was considered. As a result of axial loading, wing cracks can extend from the tips of the preexisting crack that eventually become parallel to the specimen free surface. Holzhausen and Johnson  examined several possibilities in uniaxial compressive testing of rock which result in induced lateral tensile stresses and axial splitting. The envisioned mechanisms included buckling of vertical rock slices due to presence of internal flaws, induced axial cracks due to imperfection in the specimen geometry, sliding on an inclined internal crack which can result in tensile cracking, and development of axial cracks due to presence of elliptical holes parallel to a free surface. Dey and Wang  considered the stress inhomogeneity and cracks interactions as the cause of rock facture under compressive loading. Bazant and Xiang  introduced a fracture mechanics based model for the compression failure of quasi-brittle materials. In their model, a pre-existing crack was considered to be parallel or inclined with respect to the specimen axis.
Defining petrophysical and mechanical properties of target and barrier zones are key components of the hydraulic fracture modeling process; subsequently, the selection of the detail necessary to accurately model fracture/reservoir performance is challenging. This work investigates whether using detailed petrophysical and mechanical properties provides fracture design parameters that better represent actual fracture behavior and subsequent well performance than a single-layered model.
The approach was to model an existing hydraulic fracture treatment and well performance from a well located in the northern Delaware Basin producing from the lower Brushy Canyon Formation. Models varied from a single layer model with simple-averaged, petrophysical properties to a fine resolution 1-ft model with detailed petrophysical values. Detailed core descriptions were constructed to appropriately represent the thin-bedded and micro-laminated sandstones and siltstones.
In addition, point load tests measured values of fracture toughness for specific lithofacies from 600 to 1100 psi-in½. In comparison, the default value for a sandstone system is 1000 psi-in½. Other mechanical properties, e.g., Poisson’s ratio and Young’s modulus were derived from well logs, and were within typical values.
For the fracture modeling phase, the actual treatment volumes, rates and pressures were inputted into the model along with the measured petrophysical and mechanical properties. Model net pressure was matched with the actual values to verify the output. The dimensionless fracture conductivity (FCD) from the various models ranged from 4.8 to 13.6. The range depends on the variation of lithofacies included in the fine resolution models and their associated mechanical/petrophysical properties. Adding micro-laminated and bioturbated siltstones at the expense of clean sandstone in the finer resolution models resulted in higher permeability, fracture toughness and lower stress gradient.
For the production history matching phase, simulation pressures were significantly overestimated compared to actual measured bottomhole pressures for all single layer models regardless if actual or default mechanical properties were used. The overestimation reflects a threefold increase in pore volume due to the single layer values. For the finer resolution 1-ft model, the simulation pressure was significantly below measured pressure values using default mechanical properties. However, using actual mechanical properties in the 1-ft resolution model resulted in an increase in the FCD due to the decrease in fracture toughness and stress gradient input values. As a result, a very good match was obtained between simulation and actual pressures; indicating the 1-ft model with the measured mechanical properties is a good representation of the actual reservoir system.