Measurement of Width and Pressure in a Propagating Hydraulic Fracture
Abstract Measurements of width and pressure in a propagating hydraulic fracture have been made in tests conducted at the U.S. DOE's Nevada test site. This was accomplished by creating an "instrumented fracture" at a tunnel complex (at a depth of 1,400 ft [425 m]) where realistic insitu conditions prevail, particularly with respect to stress and geologic features such as natural fractures and material anisotropy. Analyses of these data show that the pressure drop along the fracture length is much larger than predicted by viscous theory, which currently is used in models, This apparently is caused by the tortuosity of the fracture path, multiple fracture strands, roughness, and sharp path, multiple fracture strands, roughness, and sharp turns (corners) in the flow path resulting from natural fractures and rock property variations. It suggests that fracture design models need to be updated to include a more realistic friction factor so that fracture lengths are not overestimated. Introduction Hydraulic fracturing, which has proved a valuable well-stimulation technique for low-permeability reservoirs, has been the subject of considerable study for nearly 30 years. Many theories have been advanced to model the process and aid in the design of the treatment. In process and aid in the design of the treatment. In general these theories differ mainly in the approach used to model the rock deformation (i.e., the width equation). The fluid mechanics model in all cases is based on A pressure drop that is derived theoretically for parallel pressure drop that is derived theoretically for parallel flow between smooth plates or in smooth pipes (at least for laminar flow, which prevails in the large majority of fracture treatments). Attempts to verify these models have been generally limited tolaboratory studies, such as those of Blot et al., which are difficult to perform and may be impossible to scale if rock is used,postfracture well testing or production history matching analyses to deduce fracture length (e.g., those of Holditch and Lee ),analyses of fracturing pressure records by Nolte and Smith, 15 andwellbore width measurements by Smith. 16 The data from these studies are very limited and it is difficult to arrive at a consensus on the validity of the previously mentioned models. However, well testing and production history matching studies usually show that fracture lengths are overestimated considerably. This study is an initial attempt to measure pressure and width in propagating hydraulic fractures under conditions that avoid some of the size and scaling problems of laboratory tests and yet provide greater accessibility and instrumentation than field tests. These experiments were conducted at the U.S. DOE's Nevada test site, where hydraulic fractures were created and monitored from an existing tunnel complex. This initial experiment was conducted to determine whether it was feasible to measure important fracture parameters accurately and obtain significant information about fracture growth processes. Of particular importance was the pressure processes. Of particular importance was the pressure drop along the length of the propagating fracture. Background Hydraulic fractures are not the smooth parallel plates that they usually are modeled to be. Mineback experiments 17–20 have shown that there is considerable surface roughness and waviness, common en echelon fracturing and multiple stranding, and significant offsets when natural fractures are intersected. Natural fractures in core show many of these same characteristics, although the fracturing mechanism admittedly may be different. Laboratory experiments also show many of these same effects. Lamont and Jessen demonstrated the offset of hydraulic fractures at natural joints and showed the surface waviness and roughness of the fracture. Blot et al. 13 found that the roughness of the fracture surface depended on rock type and decreased with increasing confining stresses. Smith 16 measured fracture width at the wellbore with a TV camera and observed consider-able width variation or large-scale roughness. The effect of such variability of the fracture shape, path, and surface features must be an increase in pressure drop along the length of the fracture compared with that of the ideal case. This may have a significant influence on the resultant widths, lengths, and heights of the induced fracture. In the ideal case, the pressure drop for laminar flow usually is represented by a friction factor, (1) where NRe is the Reynolds number and C depends on the geometry. Huitt, Rothfus and Monrad, Rothfus et al. and Whan and Rothfus describe correlations for flow through parallel plates and tubes for both laminar and turbulent flow. For relatively smooth tubes C is 16 and for smooth parallel plates it is. Elliptic cross sections of zero ellipticity are calculated to have a C value of 2 pi 2. A generally held belief from all these studies is that in the laminar regime (NRe less than 2,000), flow through parallel plates is independent of roughness. parallel plates is independent of roughness. SPEJ P. 46