Professional Documents
Culture Documents
McLaury
e-mail: brenton-mclaury@utulsa.edu
Quamrul H. Mazumder
the Effects on Sand Erosion
Department of Computer Science,
Predicting erosion resulting from the impact of solid particles such as sand is a difficult
Engineering Science and Physics,
task, since it is dependent on so many factors. The difficulty is compounded if the par-
University of Michigan-Flint,
ticles are entrained in multiphase flow. Researchers have developed models to predict
213E Murchie Science Building,
erosion resulting from solid particles in multiphase flow that account for a variety of fac-
303 East Kearsley Street,
tors. However, no model currently accounts for the flow orientation on the severity of
Flint, MI 48502
erosion. This work provides three sets of experimental results that demonstrate pipe ori-
e-mail: qmazumde@umflint.edu
entation can have a significant impact on the amount of erosion for annular flow. A semi-
mechanistic model to predict erosion in annular flow is also outlined that accounts for
Gerardo Santos the upstream flow orientation. [DOI: 10.1115/1.4004264]
Ecopetrol, Corrosion and
Materials Engineering, Keywords: annular flow, erosion, sand
A.A. 4185,
Bucaramanga, Columbia
e-mail: gsantos@ecopetrol.com.co
1 Background and Introduction anticipated during production have been addressed [2,3] and this
has led many producers to seek out other models intended to pre-
Many oil and gas producers are forced to address issues result-
dict solid particle erosion. Recognizing the need for models to
ing from the production of formation sand along with oil and gas.
replace API RP 14E, Salama and Venkatesh [4] and Svedeman
The produced sand causes several problems such as potential
and Arnold [5] presented models to calculate an erosional thresh-
damage to the reservoir, sand accumulation in lines or equipment,
old velocity. Both of these models accounted for the amount of
sand separation issues, and erosion. Erosion problems result from
sand flowing in the system. The model of Salama and Venkatesh
sand impacting the surfaces of lines and equipment removing pipe
was based on data of Weiner [6] and assumed the particles
wall material. The material removal can result from several differ-
impacted the surface at velocities similar to the flow velocity.
ent mechanisms including cutting, ploughing, fatigue, and brittle
This is approximately valid for gases at relatively low pressures.
fracture. For all mechanisms the severity of erosion depends on
However, this model did account for the pipe diameter. The model
many factors including wall material, type and size of geometry,
of Svedeman and Arnold [5] also accounts for the size and type of
flow regime, fluid properties, and particle properties and size.
fitting. This model relies on an empirical constant derived from
Regardless of the mechanism, erosion is a serious concern. Ero-
experimental data presented by Bourgoyne [7]. These two proce-
sion can result in failure of equipment posing a danger to person-
dures are based on the experimental data and are only valid for a
nel and the environment and also result in economic losses from
limited range of conditions. For example, these procedures are not
production losses and repairing or replacing the eroded compo-
general and should not be used for a broad range of fluid proper-
nent. To alleviate the risk of unforeseen failures from sand ero-
ties. The models also do not account for particle properties such
sion, design engineers need a tool capable of predicting erosion
as size.
rates in a given geometry exposed to a given set of operating con-
Shirazi et al. [8] developed a model to predict maximum ero-
ditions. With this tool, the production rate could be optimized
sion that accounted for geometry type, size, and material; fluid
allowing the largest production rate that would produce only an
properties and rate; and sand size, shape, and density. The model
allowable amount of erosion. However, developing this tool is dif-
is based on Eq. (1) and was originally developed for use with a
ficult since the severity of erosion depends on many factors.
single-phase carrier fluid.
For years, oil and gas producers used the American Petroleum
Institute Recommended Practice 14E, API RP 14E [1], to calcu-
WV1:73
L
late a threshold erosional flowstream velocity. However, this h ¼ FM FS FP (1)
guideline does not consider sand nor does it account for many fac- ðD=Do Þ2
tors that affect erosion. Speculation on the origin of the practice
has been discussed, but the form demonstrates that it was not where h ¼ penetration rate, FM, FS ¼ empirical factors for material
intended for use in the presence of sand [2]. Misuse of the practice and sand sharpness, FP ¼ penetration factor, W ¼ sand production
resulted from its ease of application and lack of other available rate, VL ¼ characteristic particle impact velocity, D ¼ pipe diame-
methods. The shortcomings of API RP 14E for use when sand is ter, and Do ¼ reference pipe diameter, 1 in. (25.4 mm)
Values for FM are determined experimentally by direct
impingement erosion testing of the material of interest. Values for
Contributed by the Petroleum Division of ASME for publication in the JOURNAL
OFENERGY RESOURCES TECHNOLOGY. Manuscript received June 30, 2009; final
FS have also been determined experimentally and range from 0.2
manuscript received May 5, 2011; published online June 27, 2011. Assoc. Editor: for spherical particles (glass beads) to 1.0 for sharp, angular par-
Andrew K. Wojtanowicz. ticles. FP is a penetration factor that is a function of geometry.
Mazumder [11] performed erosion experiments under very sim- 2.3 Acoustic Sand Monitor Testing. Viswanathan [19] also
ilar conditions to Santos. In fact, the same 1-in. experimental fa- performed work examining the effect of upstream flow orientation
cility including test cell was used for both studies. The same type for multiphase flow. Instead of examining erosion, Viswanathan
of sand with an average diameter of 150 lm was also used. How- investigated acoustic sand monitor performance. A commercially
ever, Mazumder used a stainless steel erosion specimen instead of available acoustic monitor was utilized for this work. The monitor
an aluminum specimen. is a passive device that “listens” for acoustic signals generated by
flow noise and particles striking the walls. In an effort to reduce
2.2 Sand Distribution Measurements. Since the results the effect of flow noise, the transducer signal for a given fre-
obtained for the erosion testing present a phenomenon that is not quency range associated with particle impacts is evaluated. The
documented as will be discussed in Sec. 3, additional studies were transducer signal is converted to a voltage that is reported in real
proposed to gain insight to the reason behind the differences. The time. Calibration of the monitor is necessary if it is desired to
next set of experiments was to gather information on the sand dis- determine the sand rate from the monitor output. However in this
tribution in the upstream flow. Intrusive probes were placed work, calibration is not necessary since the goal is simply to com-
upstream of the horizontal and vertical test sections of the facility pare the acoustic monitor output for horizontal and vertical orien-
described in Sec. 2.1. The Pitot-type probes were able to traverse tations for fixed flow conditions and sand rates. He examined
most of the inner diameter of the pipe. A schematic of the probe annular, slug, bubble, and dispersed bubble flows as well as sin-
and the five locations, where samples of liquid and sand were col- gle-phase gas and liquid. However, this work will only focus on
lected are shown in Fig. 4. The samples were collected for the test the annular flow cases. The output created by the acoustic sensor
conditions used for the erosion experiments: a fixed superficial is a function of the number and size of particles impacting the sur-
liquid velocity and three superficial gas velocities. A fixed quan- face and the impacting velocity of the particles. In this sense, a
tity of sand and liquid were used for each test, and once again the correlation between the acoustic monitor output and measured
concentration of the sand was approximately 2% by mass of the erosion is expected.
liquid. The probe collected the mixture of water and sand during A schematic of the 2-in. diameter facility used for acoustic
the entire test. More details on the testing procedure are described monitor testing is shown in Fig. 5. The water and sand from the
by Santos [18]. slurry mixing tank are injected at the base of the boom, and for
the annular flow cases considered, no additional water was
required from the liquid tank. A commercially available acoustic
sensor was placed downstream of a metallic elbow. The elbow
was at the end of a 7.6 m (25 ft) pipe placed on a boom that could
Fig. 2 Test cell used for erosion experiments Fig. 3 Sand size distribution
be raised from the horizontal to the vertical position. A photo- factors of 1.9, 4.7, and 2.7 for the superficial gas velocities of 34,
graph of the boom is shown in Fig. 6. Most of the piping on the 27, and 19 m=s, respectively.
boom was clear PVC, so that visualization of the flow pattern Mazumder also used a superficial liquid velocity of 0.30 m=s
could be performed. The sand used during this testing was similar and three superficial gas velocities of 34, 27, and 19 m=s (112, 90,
to the sand used by Santos and Mazumder. and 62 ft=s) to conduct erosion testing on stainless steel speci-
mens. Figures 10–12 show the mass loss versus sand throughput
3 Experimental Results for the three gas velocities for both horizontal and vertical orienta-
tions. The results demonstrate similar results as gathered by San-
3.1 Erosion Testing. Erosion experiments were performed tos. It is interesting to note that the erosion is higher for the
by Santos using aluminium specimens at a superficial liquid ve- vertical orientations by factors of approximately 2.5, 4.9, and 2.6
locity, Vsl, of 0.30 m=s (1.0 ft=s) and for superficial gas velocities, for the superficial gas velocities of 34, 27, and 19 m=s. These fac-
Vsg, of 34 m=s (112 ft=s), 27 m=s (90 ft=s), and 19 m=s (62 ft=s). tors are very similar to those presented previously by Santos.
Additional details of the testing are provided in Sec. 2.1. The
amount of erosion was quantified by measuring the mass loss of
the erosion specimen. Figures 7–9 show the mass losses of the 3.2 Sand Distribution Measurements. The next sets of data
horizontal and vertical specimens as a function of the sand were collected to aid in the understanding of the effects of pipe
throughput for the superficial gas velocities of 34, 27, and 19 m=s, orientation on the erosion results. The following results are pre-
respectively. For each test condition, the erosion in the vertical sented as percentages of the water and sand collected by the probe
section was higher than in the horizontal section. Based on the av- at a given location of the total mass of liquid and sand used for
erage erosion, the erosion was higher in the vertical specimens by the test. Figures 13–15 show the percentage of liquid and sand for
Vsl ¼ 0.30 m=s and Vsg ¼ 34, 27, and 19 m=s, respectively, for the
vertical orientation. These figures demonstrate the distributions of
the liquid and sand are relatively uniform across the pipe. The fig-
ures may indicate that the amount of water and sand collected in
the center of the pipe decreases with decreasing gas rate. (The
Fig. 5 Schematic of 2-in. multiphase erosion test facility Fig. 6 Photograph of boom for 2-in. erosion test facility
Fig. 9 Comparison of erosion for horizontal and vertical test Fig. 12 Comparison of erosion for horizontal and vertical test
sections for Vsg 5 19 m=s and Vsl 5 0.30 m=s, aluminum sections for Vsg 5 19 m=s and Vsl 5 0.30 m=s, stainless steel
specimen specimen
Fig. 15 Fraction of liquid and sand by location, Vsg 5 19 m=s and Vsl 5 0.30 m=s, vertical orientation
the gas core travelled at a velocity 80% of the gas velocity; this is can be used for horizontal and vertical upstream flow. This model
substantially higher than the average liquid film velocity. As the is an extension of the single-phase model proposed by Shirazi et
sand rate increases, the monitor output will remain approximately al. [8] discussed in detail in Sec. 1. There is some similarity to
the same until the sand impacts increase the noise level above the models proposed by Viswanathan [19] and Mazumder [11].
background value. The point at which the noise level increases The model assumes that particles are present in the gas core
above the background noise is referred to as the threshold sand and the liquid film of annular flow. The particles are assumed to
rate that can be detected by the acoustic sand detector. As the travel in the same manner as the liquid; meaning if a certain frac-
sand rate increases beyond the threshold value, the monitor output tion of liquid is in the gas core in the form of droplets (entrain-
varies fairly linearly on a log–log plot with sand rate. Figure 20 ment fraction), then an equal fraction of the particles is also in the
show the results for two additional annular flow conditions for gas core. So, a liquid entrainment relation is used to determine the
both orientations along with the data presented in Fig. 19. The fraction of sand in the gas core and in the liquid film. The entrain-
results for each flow orientation are very similar with the vertical ment model used for this study is provided by Ishii and Mishima
orientation being consistently higher than the same condition for [22]. More discussion on the selection of this model is provided
the horizontal orientation. The results do show that the lowest gas by Mazumder [11]. For this work, the entrainment is assumed to
rate produces slightly lower monitor output. Figure 20 demon- be equal for the vertical and horizontal cases for a given set of
strates that the acoustic monitor output for the vertical section is operating conditions. Once the split of particles between the gas
consistently higher than for the horizontal section for all three core and liquid film is known, the next step is to determine the
conditions considered. Even if the effect of the background noise representative impact velocity of particles in each region. An im-
is removed, the adjusted output is still higher for the vertical ori- portant factor affecting the impact velocity of the particles is the
entation. This indicates the particle impact momentum is less in liquid film thickness. A mechanistic model proposed by Paz and
the horizontal case. This can mean that either fewer particles are Shoham [23] is used to determine the film thickness for both the
striking the wall near the elbow test section and=or that the par- vertical and horizontal orientations. The film thickness for the ver-
ticles are impacting the walls at lower velocities for the horizontal tical orientation is assumed to be symmetric around the pipe. The
section. film thickness at the top and bottom of the pipe are calculated for
the horizontal orientation. The film thickness at the bottom of the
pipe is used for the semimechanistic model being proposed in this
4 Semi-Mechanistic Erosion Model to Account for work for horizontal flow.
Flow Orientation A two-step approach is used to determine the representative
No erosion prediction model is currently available that accounts impact velocity of particles present in the gas core. Figure 21 is a
for the upstream pipe orientation. This section presents a model schematic representing the approach used. The motion of a repre-
that predicts erosion in elbows for annular flow conditions that sentative particle is simplified to a one-dimensional path. A
Fig. 17 Fraction of liquid and sand by location, Vsg 5 27 m=s Fig. 18 Fraction of liquid and sand by location, Vsg 5 19 m=s
and Vsl 5 0.30 m=s, horizontal orientation and Vsl 5 0.30 m=s, horizontal orientation
Fig. 20 Normalized average acoustic monitor output for annular flow (three conditions)