A Thermally Actuated Gas-Lift

Safety Valve
Eric Gilbertson and Franz Hover, Massachusetts Institute of Technology, and Bryan Freeman, Chevron ETC
Summary
Most offshore oil wells use ow-control safety valves that are ei-
ther actively controlled from the surface or passively actuated by
pressure changes. In this paper, we propose a passive, thermally-
actuated safety valve that adds redundancy to well systems and
signicantly increases well safety. We use quasistatic thermal
modeling, validated with experimental downhole data, to identify
and characterize a suitable temperature regime for shape memory
alloy actuation. We develop a complete loading and shut-in proce-
dure with the new valve that is consistent with current practice in
articial lift. Finally, through a true-scale stainless-steel proto-
type, we demonstrate essential manufacturability and successful
valve closure in accordance with predictions.
Introduction
Flow-control safety valves are critical in offshore oil and gas pro-
duction to ensure that produced hydrocarbons do not reach the
surface or seaoor uncontrollably. Currently, most safety valves
are either actively controlled from the surface or are passively
actuated through pressure or ow-rate changes. Safety valves are
typically located in four places in the well: at the wellhead in a
blowout preventer (on the seaoor), downhole in the well tubing
[subsurface safety valves (SSV)], in the annulus [annular safety
valves (ASV) in gas-lifted wells], and between the annulus and
tubing [gas lift valves (GLV) in gas-lifted wells]. Valves in the
annulus and between the annulus and tubing can be either above
or below the seaoor.
The wellhead blowout preventer is a mechanical device
composed of multiple shear rams and annular elastomer wedge
valves (Holand and Rausand 1987). The blowout preventer is
used during drilling, workover, and remedial operations, and
therefore is only installed on the well for a very small portion of
the wells life. Blowout preventers are most commonly controlled
by a hydraulic or electric connection to the surface (Ma et al.
2006), with backup control possible using acoustic channels or
remotely-operated vehicles (ROVs).
SSVs are most commonly apper or ball valves actuated by
electric (Tanju and Worman 2009) or hydraulic power from the
surface (Bellarby 2009). Acoustic signals are also used to commu-
nicate with some downhole sensors (Prensky 2006) and this may
someday be applied to downhole safety valves.
ASVs may be used in deepwater wells (deeper than 6,000 ft)
to prevent hydrocarbons from owing uncontrollably up the annu-
lus (Bellarby 2009). Like SSSVs, ASVs are controlled from the
surface by hydraulic control lines, which hold the valves open
during normal operation and allow the valve to close if hydraulic
pressure is lost.
Most GLVs have a bellows, stem and ball, as well as a check
valve to prevent backow of hydrocarbons up the annulus (Takacs
2005). The check valve is passively actuated by a pressure differ-
ential, and the bellows and ball close if the annulus pressure drops.
In this paper, we contend that a new type of ow control valve
(thermally-actuated) is a viable alternative that would add redun-
dancy and signicantly increase safety. Our main contributions
are as follows. We used quasistatic thermal modeling, validated
with experimental downhole data, to identify and characterize a
suitable temperature regime for shape-memory-alloy actuation in
gas-lifted wells. Second, we have developed a complete loading
and shut-in procedure with the new valve that is consistent with
current practice in articial lift. Finally, through a true-scale
stainless-steel prototype, we have demonstrated essential manu-
facturability and successful valve closure in accordance with
predictions.
Artificial Gas Lift and the NewConcept
Gas lift is an articial lifting method used to produce oil from
wells that do not ow naturally and to accelerate production rate
in naturally owing wells. Natural gas is injected through the well
annulus and into the well tubing at a downwell location (as shown
in Fig. 1). The gas mixes with the produced uids in the tubing,
aerating the produced uids and causing them to rise to the sur-
face (Brown 1980).
GLVs are designed to be one-way valves that allow gas to pass
from the annulus to the tubing, but prevent reservoir uids from
passing through to the annulus (Brown 1980). Most valves contain
a pressurized bellows and an internal check valve (see Fig. 2). In
an injection-pressure-operated valve (IPO), the bellows opens
when the injection gas is pressurized above a threshold value
(Takacs 2005). A GLV fails if it allows uid passage from the
tubing to the annulus (Carlsen et al. 2010). Two main criteria are
sufcient for failure in an IPO valve: (1) the reverse-ow check
valve leaks and (2) a combination of high tubing pressure and low
gas pressure allows the bellows valve to open (see Fig. 2). (It is
generally difcult to meet the second of the criteria, especially
with a large bellows and small port.) The pressure relation that
satises these criteria is described by the equation
P
open
< P
ann
< P
tube
; 1
where P
open
is the annulus pressure required to open the bellows
valve, P
ann
is the actual pressure in the annulus, and P
tube
is the
pressure in the tubing. GLVs indeed do fail in practice, with the
most common failure mode a failed check valve, followed by an
eroded ball/seat interface, and a failed bellows. Our goal is to aug-
ment the basic GLV with an additional, thermally actuated safety
valve.
Thermally actuated safety valves have been used since the
1930s in many different applications: internal combustion engines
(Bible 1977), gas-burning stoves (Place 1979; Pymm 1935; May-
eld and Haselsweidt 1982), steam traps (Tarvis 1984), smoke
detectors (Kelly and Fredd 1972), and anti-scald water valves
(Hart and Williams 1980; Homma 1990). For actuation, the valves
use bimetallic strips (Tarvis 1984), gas- or liquid-lled bellows
and diaphragms (Place 1979; Tarvis 1984), expanding springs
(Hart and Williams 1980), dissolving solids (Kelly and Fredd
1972), and shape-memory alloy springs and wires (Hart and Wil-
liams 1980; Homma 1990). To the authors knowledge, there are
no widely-used thermally actuated safety valves in the offshore
oil industry, however. A thermally actuated safety valve would
rely on a change in temperature between normal operation and an
undesirable ow event. For example, as we show in this paper, in
gas-lifted wells a reservoir uid backow event would lead to a
rise in GLV temperature (because injection gas is cooler than res-
ervoir uid), and this signal is sufcient to trigger the safety valve
to close. By a similar argument, annular safety valves and
. . . . . . . . . . . . . . . . . . . . . . . .
Copyright VC 2013 Society of Petroleum Engineers
Original SPE manuscript received for review 17 August 2011. Revised manuscript received
for review 6 February 2012. Paper SPE 161930 peer approved 9 April 2012.
February 2013 SPE Production & Operations 77
blowout preventers could also be actuated by thermal signals if
undesirable ow events cause signicant temperature changes.
A thermally actuated valve has the advantage of requiring no
surface communication to actuate, thus reducing actuation time
for the valve and increasing the likelihood that it will still work if
something catastrophic happens on the surface rig. The valve is
also potentially much cheaper than electrically or hydraulically
actuated alternatives, which would require thousands of feet of
communication line in deepwater wells. Moreover, it would add
redundancy to any ow-, pressure-, or actively controlled safety
system by providing an independent signal for valve closure.
Steady-State Thermal Model
In this section, we derive a model for the steady-state oil and gas
temperature proles in the gas-lifted well. This model is fairly
simple, but reects techniques used by the industry that have been
shown effective (Sagar et al. 1991). We use model results to dem-
onstrate feasibility of a thermally actuated safety valve.
Steady-State Assumptions. Piping. For the piping, we assume
10-cm-thick cement insulation around the annulus and 1-cm-thick
steel tubing and annulus pipes (Brown 1980).
Mixture. For the gas/oil mixture, we assume heat loss to the
annulus through convection and conduction; and heat capacity
and conductivity are weighted averages of gas and oil properties:
c
mix

_ m
gas
_ m
mix
c
gas

_ m
oil
_ m
mix
c
oil
2
k
mix

_ m
gas
_ m
mix
k
gas

_ m
oil
_ m
mix
k
oil
; 3
where c is capacity, k is conductivity, and _ m is mass ow rate.
We also assume that mixture temperature just above the injec-
tion point is a weighted average of gas and oil temperatures
T
mix

c
gas
_ m
gas
c
mix
_ m
mix
T
gas

c
oil
_ m
oil
c
mix
_ m
mix
T
oil
: 4
Gas. For the injection gas we assume gas-formation and gas-
tubing heat transfers by convection and conduction; cooling from
water for the section of well subsea but above ground; a surface
gas temperature of 300K; and a constant gas mass ow rate into
the well. The assumption of constant gas mass ow rate is made
to linearize the differential equations of the system.
The gas mass ow rate is calculated by
_ m
gas

q
gas
q
oil
b _ m
oil
; 5
where q
gas
is the gas density, q
oil
is the oil density, and b is the
gas/oil volumetric ratio at injection depth. The gas density at the
surface can be calculated using the ideal gas law
q
gas

P
gas
M
gas
RT
gassurf
; 6
where P
gas
is the gas pressure at the surface, M
gas
is the gas molar
mass, R is the ideal gas constant, and T
gassurf
is the surface gas
temperature. These parameters ultimately determine the gas mass
ow rate at the surface, and this same mass ow rate is then used
for all depths.
Modeling Approach. In this model, two coupled differential
equations are derived for the temperature proles in the annulus
and tubing.
Annulus. We rst look at the heat transfer in a differential
control volume in the annulus. In Fig. 3, a differential control vol-
ume is drawn around a section of the well annulus between the
depths of x and x dx. For simplicity, the control volume of the
half of the annular segment is represented in 2D as a rectangle,
where r
case
is the casing inner radius, r
tube
is the tubing inner ra-
dius, r
case
r
tube
is the width of the differential element, and dx is
the height of the differential element. Heat is transferred into the
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
Bellows
Valve
Injection
Gas
Venturi Orifice
Check Valve
Fig. 2GLV model. Normal operation (left, gas injected though
GLV) and failure (right, oil owing wrong direction through
GLV).
Q
tube
T
tube
T
gas
(x+dx)
T
gas
(x)
b
X
X+dX
T
form
Q
gasout
.
Q
gasin
.
Q
form
.
.
Fig. 3Heat transfer model for annulus control volume. Heat
transferred from tubing, formation, and gas ow.
Reservoir
GLV
Tubing
Annulus
O(5 km)
Fig. 1Schematic of oil well with gas lift valve (GLV). Top of g-
ure represents seaoor. Gas is injected into the annulus,
passes through the GLV, and into the tubing, allowing product
to reach the surface.
78 February 2013 SPE Production & Operations
control volume by convection and conduction through the casing
wall, convection and conduction through the tubing wall, and
mass ow into the top of the control volume.
Heat is transferred out of the control volume through mass
ow out the bottom of the volume. Heat transfer from the casing
to the gas is given by
_
Q
form
x

T
form
x T
gas
x

R
totg
; 7
where
_
Q
form
is the heat transfer from the casing to the control vol-
ume, T
form
(x) is the formation temperature at depth x, and T
gas
(x)
is the gas temperature at depth x. The formation temperature in
Kelvins is assumed to be of the form
T
form
x 273
1
40
x; 8
where x is measured in meters below the seaoor (Fridleifsson
et al. 2008). R
totg
is the total thermal resistance across the casing
interface, which is the sum of the conduction and convection
resistances:
R
totg

b
k
cem
2pr
case
dx

1
h
form
2pr
case
dx
; 9
where b is the casing wall thickness, k
cem
is the cement thermal
conductivity, and h
form
is the convective heat -transfer coefcient
of the casing wall.
The convective heat transfer coefcient is given by
h
form
Nu
form
k
gas
2r
case
; 10
where Nu
form
is the formation Nusselt number (the dimensionless
ratio of convective to conductive heat transfer). Heat transfer
from the tubing to the gas is given by
_
Q
tube
x

T
tube
x T
gas
x

R
tott
; 11
where
_
Q
tube
is the heat transfer from the tubing to the annulus,
and T
tube
(x) is the tubing temperature at depth x. R
tott
is the total
thermal resistance across the tubing interface, which is the sum of
the conduction and convection resistances:
R
tott

a
k
tube
2pr
tube
dx

1
h
tube
2pr
tube
dx
; 12
where a is the tubing wall thickness, k
tube
is the tubing-wall ther-
mal conductivity, and h
tube
is the convective heat transfer coef-
cient of the tubing wall given by
h
tube
Nu
tube
k
mix
2r
tube
; 13
where Nu
tube
is the tubing Nusselt number. Heat transfer through
the control volume caused by mass ow is given by
_
Q
gasout
x
_
Q
gasin
x
_ m
gas
c
gas

T
gas
x dx T
gas
x

;
14
where
_
Q
gasout
is the heat transfer out the bottom of the control vol-
ume caused by mass ow and
_
Q
gasin
is the heat transfer into the
control volume caused by mass transfer.
An energy balance for the control volume yields the equation
_
Q
gasout
x
_
Q
gasin
x
_
Q
tube
x
_
Q
form
x: 15
Combining Eqs. 14 and 15 and dividing both sides by dx yields
the differential equation
_ m
gas
c
gas
dT
gas
x
dx

T
tube
x T
gas
x
R
tott

T
form
x T
gas
x
R
totg
:
16
Tubing. Using similar methods, a differential equation can be
derived for a control volume in the tubing, and the result is
_ m
mix
c
mix
dT
tube
x
dx

T
tube
x T
gas
x
R
t
: 17
We employ a shooting method to solve the equations, where T
tube
and T
gas
at the injection site are selected so that boundary condi-
tions at the surface are met.
Comparison With Experimental Data. To check the validity of
the model, temperature proles were compared with data from an
actual well. A temperature survey from a land-based gas-lifted
well was provided by the project sponsor, where data is for the
temperature inside the tubing. Of the 20 parameters required for
the model, eight were provided and 12 were standard values
looked up in other sources (such as oil-specic heat and oil ther-
mal conductivity) (see Table 1). The only unknown parameters
were the Nusselt numbers of the well.
Fig. 4 shows the steady-state tubing temperature prole for a
tubing Nusselt number of 700 and formation Nusselt number of
1,000. Well parameter values were allowed to vary by within 10%
of a given value when tting model to data. Model 1 was chosen
as the best t, with several other candidate lines also shown. The
temperature difference at injection depth is considered typical on
the basis of the three different well data sets provided.
Thermal Properties of Entire System. The temperature model
was then expanded to describe a subsea well, with an exponential
model assumed for ocean temperature vs. depth (Felzenbaum
1992) and an ocean depth of 1000 m assumed. The same well pa-
rameter values were used as in the land-based well data. The well
injection depth is 3000 m below the seabottom. Resulting plots
with and without the tubing subsurface safety valve closed and
Nusselt numbers 700 and 1,000 are given in Figs. 5 and 6, respec-
tively. Also plotted are ideal shape-memory alloy transition
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . .
. . . . . . . . . . .
TABLE 1PARAMETER VALUES USED IN MODEL-DATA
COMPARISON, MODEL 1
Parameter Value Units Source
w 2500 m Well Data
_ m
mix
24 kg/s Well Data
_ m
gas
16.5 kg/s Well Data
c
gas
2775 J/kg-K (Das et al. 2007)
a 0.0085 m (Brown 1980)
k
tube
50 W/m-K (Ashby 1996)
r
tube
0.0457 m Well Data
b 0.09 m (Brown 1980)
k
cem
1.73 W/m-K (Ashby 1996)
r
case
0.1016 m Well Data
c
oil
1841 J/kg-K (Das et al. 2007)
k
oil
0.15 W/m-K (Das et al. 2007)
k
gas
0.04 W/m-K (Das et al. 2007)
T
res
356 K Well Data
T
gassurf
300 K Well Data
fA
`
45 Unitless Well Data
r
oil
900 kg/m
3
(Takacs 2005)
P
gas
6:4x106 Pa Well Data
_ m
oil
7.5 kg/m
3
Well Data
M
gas
0.016 kg/mol (Atkins and Jones 2005)
*Only Nusselt numbers not given in well data or other sources.
February 2013 SPE Production & Operations 79
temperatures, which will be described in detail in a later section.
The plots show that, under these conditions and assumptions,
there is a 4

temperature difference between gas and reservoir oil

temperature at injection depth with the tubing safety valve open,
and a 5

temperature difference with the tubing safety valve

closed. Although every modeled or real-world system will be dif-
ferent, we think that these temperature differences represent well
the thermal signals available downhole. If the safety valve is de-
sirable in installations with smaller temperature differentials, one
could consider modications to the design (e.g., inclusion of ther-
mal insulation) to enhance it.
Thermally Actuated Safety Valve. We now describe the design
of a thermally actuated safety valve that requires no surface com-
munication or external power sources to actuate. Power downhole
is a problem that the industry has considered through electric
lines, hydraulic lines, and even extraction of energy from the uid
ow to provide power. Surface communication is also a problem,
potentially requiring thousands of meters of communication line
in ultradeep wells. We have a big advantage on both of these
counts by using shape-memory alloy wires for thermal actuation.
Concept Details. The thermally actuated gas lift safety valve
is a trunnion ball valve with cylindrical side extensions located in
the top section of the venturi orice, just below the bellows valve
(see Figs. 7 and 8). A ball valve was chosen because its closure is
not affected by a pressure difference across the valve (Smith and
Zappe 2004). Also, in the open position, the ball valve does not
affect the injection gas ow, and should be subjected to less wear
than the existing check valve. Corrosion may occur if the valve is
partially open, and this would be a major engineering question to
study as the technology is developed. It is not yet known how
small errors in ball-valve position affect gas-ow rate, but ball
valves are already used as downhole safety valves in the tubing of
some wells (Bellarby 2009) and are accepted in the offshore oil
industry. The ball-valve side extensions have small pins press t in
perpendicularly which constrain the valve motion to 90

of rota-
tion between vertical (fully open) and horizontal (fully closed).
Ball bearings support the side extensions on the outside of the
stoppers, and O-rings form a valve seal.
The ball valve is designed to rotate 90

between the open and

closed positions, though it actually stops ow at a rotation angle
270 280 290 300 310 320 330
Temperature (K)
Normal Operation
Ocean Surface
Gas
Gas-Oil Mixture
Ground, Water
Pure Oil
Ocean Bottom
Injection Depth
M
f
M
s
A
s
A
f
D
i
s
t
a
n
c
e

u
p

F
r
o
m

I
n
j
e
c
t
i
o
n

(
m
)
1000
0
1000
2000
3000
4000
5000
340 350 360
Fig. 5Steady-state tubing and annulus temperature proles
with tubing safety valve open. Direction of ow is shown for
gas, oil, and mixture. Ideal shape-memory alloy transition tem-
peratures shown relative to gas and oil temperatures at injec-
tion depth.
270
1000
0
1000
2000
3000
4000
5000
280
Ocean Surface
Unloading
Ocean Bottom
Injection Depth
M
f
M
s
A
s
A
f
Gas
Gas-Oil Mixture
Ground, Water
Pure Oil
290 300 310 320 330 340 350 360
Temperature (K)
D
i
s
t
a
n
c
e

u
p

F
r
o
m

I
n
j
e
c
t
i
o
n

(
m
)
Fig. 6Steady-state tubing and annulus temperature proles
with tubing safety valve closed and gas passing through the
bottom GLV at the end of an unloading procedure. Direction of
ow is shown for gas, oil, and mixture. Ideal shape-memory
alloy transition temperatures shown relative to gas and oil tem-
peratures at injection depth.
Stopper
Torsion
Spring
Torsion
Spring
Bearings
O-Rings
SMA
Wire
SMA
Wire
Fig. 7Ball valve diagram. SMA wires contract to rotate valve
closed. Wires expand and torsion spring rotates valve open.
320 325 330 335 340 345 350 360
Temperature (K)
D
i
s
t
a
n
c
e

u
p

F
r
o
m

W
e
l
l

(
m
)
1000
500
500
1000
1500
2000
Ground Level
Injection Depth
Model 1
Model 2
Model 3
Model 4
Data
2500
0
355
Fig. 4Comparison between model and data for steady-state
tubing temperature vs. depth. Good model-data agreement in
Model 1, for tubing Nusselt number of 700 and formation Nus-
selt number 1,000.
80 February 2013 SPE Production & Operations
less than 90

, as shown in Fig. 9. The minimum closure angle is

given by h 2asin(r
b
/r
o
), where r
o
is the radius of the orice, and
r
b
is the ball-valve radius. For ball dimensions designed to t in a
standard Schlumberger XL-175 gas lift valve (r
b
6 mm, r
o
11
mm), the closure angle is 66

.
The ball valve is actuated by shape-memory alloy wires which
are tied to the ball-valve side extension, wrapped one-half revolu-
tion around the side extension, and attached to the GLV below the
side extension (see Fig. 7). Shape-memory alloys undergo a solid-
state phase change between a Martensitic low-temperature state
and an Austenitic high-temperature state when heated or cooled.
These alloys are said to have memory because they return to the
same low-temperature shape whenever cooled to the Martensitic
state and to the same high-temperature shape when heated to the
Austenitic state. Shape-memory alloys are generally used in either
spring or wire form for actuation, and for this design the wire
form best ts into the existing GLV geometry.
Shape-memory alloys do not have a single transition tempera-
ture between Martensite and Austenite, but instead undergo a
hysteresis with different transition temperatures depending on
whether the alloy is being cooled or heated. This hysteresis is
shown schematically in Fig. 10, which plots strain vs. temperature.
When the alloy is being heated, A
S
represents the start of the transi-
tion from Martensite to Austenite, and A
f
represents the nal tran-
sition to Austenite. When being cooled, Ms represents the start of
the transition from Austenite to Martensite and M
f
represents the
nal transition to Martensite. In order for a thermally actuated
safety valve to be feasible using a shape-memory alloy there must
be sufcient temperature difference between the cold gas and hot
oil to actuate the SMA. Commercially available shape-memory
alloys have a minimum hysteresis temperature spread of 3

C
between A
S
and A
f
temperatures (Surbled et al. 2001), and 15

C
between A
f
and M
f
temperatures (Lagoudas 2010). Thus there must
be at least a 3

C temperature difference between gas and oil at

injection depth during normal operation, and at least a 15

C tem-
perature difference between oil and gas temperature during cool-
ing. These required temperature differences are less than the
available temperature differences of 4

C and 55

C determined
from the thermal modeling; thus, SMA actuation is feasible in the
gas-lifted oil wells modeled.
Based on the alloy proportion, a shape memory alloy can have
transition temperatures within the range of 150

C to 800

C
(Sczerzenie 2004). In this thermally actuated safety valve, the
shape memory alloy is set to have a transition temperature A
S
im-
mediately above the gas temperature and A
f
immediately below
the oil temperature (shown in Fig. 5). The M
S
and M
f
transition
temperatures are set to be between the oil temperature and the
coldest temperature attainable in the valve-cooling scenario
(described later in the operational procedures section, and in Fig.
5). If oil begins passing through the GLV, the assembly will heat
up, in turn heating up the shape-memory alloy wires to the oil res-
ervoir temperature. The wires contract as they heat up past As and
Af, and will thus pull the ball valve into the closed position. Tor-
sion springs are also wound around the ball valve side extensions.
Thus, if the shape-memory alloy cools and transitions to the Mar-
tensitic state, the wires will expand and the torsion springs will
pull the ball valve back into the open position.
Shape-memory alloys are well-suited for the harsh environ-
mental conditions of an ultradeep oil well. Nitinol, one of the
most common commercially available alloys, has corrosion resist-
ance properties similar to those of stainless steel (Deurig 1990).
Nitinol wires are known to creep under high stresses and tempera-
tures close to the melting point (Kato et al. 1999), though the
highest well temperatures are still much lower than the melting
point of Nitinol. Moreover, the design proposed in this paper
allows for some wire creep because the valve stops backow
when only rotated to 66

instead of the full 90

for which it is
designed.
Fig. 8Ball valve 3D picture. Safety valve located in venture ori-
ce section of GLV.
Valve Open Valve Closed
2r
b
r
o

Fig. 9Ball valve closure angle is less than 90

, depending on
parameters r
b
and r
o
.
Strain
Martensite
M
f
M
s
A
s
A
f
Austenite
T
e
m
p
e
r
a
t
u
r
e
Fig. 10Shape-memory alloy hysteresis. Alloy changes from
Martensite to Austenite when heated above A
S
and A
f
, and
changes back to Martensite when cooled below M
S
and M
f
.
February 2013 SPE Production & Operations 81
Operational Procedure in Gas-Lift System. As we will now
show, the thermally actuated safety valve is compatible with cur-
rent well unloading and shut-in operations. The procedure
described here assumes there is sufcient lift gas pressure to lift
from the lowermost mandrel in the well, where injection pressure
exceeds that required for lift. Well unloading is a process to
remove drilling and completion uids from a well to begin pro-
duction, and a shut-in is a temporary halt in production. Well-
startup procedures for unloading and shut-in are similar, and only
unloading will be described here.
Fig. 11 shows a schematic three-valve well and a graphical
representation of the new unloading procedures in this well. The
plot shows the changes in pressure over time in the annulus and
tubing at the three GLV depths. The bellows pressures of each
GLV (P1, P2, and P3) are plotted as horizontal lines for compari-
son, and vertical lines representing the different stages in time of
the unloading process are also plotted.
At the system level, only the GLV at operating depth contains
the thermally actuated ball valve [referred to subsequently as the
thermal valve (GLV2) in the gure]. The operating valve is the
second lowest in the tubing, with an additional standard unloading
valve located below the operating valve (GLV 3). The bellows of
the thermal valve is pressurized to a lower value (P2) than all of
the other GLV bellows in the system. During the unloading pro-
cess, the additional lower unloading valve GLV 3 is used to cool
the thermal valve. In this operational procedure, the tubing sub-
surface safety valve is closed for the duration of the unloading
process to allow better convective cooling of the thermal valve.
Initially, at time 0, the annulus and tubing are lled with kill uid
and at the same pressure. The thermal valve is at the hot steady-
state oil temperature and is closed. Annulus pressure is greater
than bellows pressure at each valve and thus the bellows valves
are all open. Gas is injected into the annulus, increasing the annu-
lus pressure and pushing uid from the annulus through the nor-
mal GLVs into the tubing. Fluid level in the annulus drops until
GLV 1 is uncovered at Time 1. Gas passes into the tubing and
tubing pressure decreases to steady values at Time 2. This allows
more annulus uid to pass through GLV 3 into the tubing until the
annulus uid level drops to the level of GLV2 at Time 3 and then
to GLV3 at Time 4. Gas now passes through both GLV 1 and
GLV 3. Tubing pressures at levels GLV2 and GLV3 drop. Annu-
lus pressure drops because gas is passing through a larger total
orice area, causing GLV 1 to close at Time 5. Until now, the pro-
cess is identical to standard unloading procedures. GLV 2 cools
by convection, the thermal valve opens, and GLV 2 starts passing
gas at Time 6. As before, annulus pressure drops. By Time 7, the
annulus pressure has dropped below P3, but is still above P2.
Thus only GLV 2, the thermal valve, is passing gas as desired.
Now the tubing safety valve is opened and normal production
begins.
The thermal valve is located approximately 15 m above the
bottom unloading valve so that it is far enough above to be
retrieved by wireline techniques with current accuracies, but not
far enough to signicantly affect well-production rates. In normal
continuous operation, gas ows through the thermal valve while
all other GLVs remain closed.
Prototypes. Two prototype safety valves were created and
tested under simulated oil well temperatures, a 3x-scale plastic
prototype and a true-scale stainless steel prototype. The purpose
of the prototypes is to demonstrate that the ball valve will com-
pletely close when the SMA wires are heated by conduction (in
this case using temperature-controlled water) above the Austenitic
transition temperature A
f
and completely open when SMA wires
are cooled below the Martensitic transition temperature M
f
. The
great majority of practical applications for SMA involve electrical
current for heating, and so it is important to characterize and
understand the implications of heating through the ambient
medium.
The prototypes are designed so that an existing GLV can be
retrotted with the thermally actuated part with minimal addi-
tional cost. To retrot a standard XL-175 GLV requires simply
cutting the venturi orice section into two pieces, milling out
pockets for the valve and wires, adding the safety valve, and bolt-
ing the two pieces back together. The tightest machining toleran-
ces required are 0.013 mm (0.0005 in.), which is attainable with
standard machine tools. In this section, we describe the true-scale
stainless-steel prototype.
The prototype housing consists of a cylinder representing the
section of the GLV between the check valve and the bellows
valve (see Fig. 12). The housing is cut horizontally into two sec-
tions, with the ball valve located between the sections. The valve
is actuated on one side by Nitinol wires. The Nitinol wire used
has a stated As transition temperature 70

C, produced by Dynal-
loy (2010). (Actual experimental As will depend on stress in the
wire and may be slightly different than 70

C). Though the Nitinol

used does not have the tight hysteresis spreads that will be needed
in the nal implementation of the valve, it is easily commercially
available and sufcient to demonstrate proof of concept prototype
testing at realistic shallow oilwell temperatures (in this case up to
70

C).
The Nitinol is attached to the ball-valve side extension with
knotted ends. The wire is then wrapped one half revolution around
Thermal
GLV
Normal
GLV
GLV 3
GLV 2
GLV 1
Flapper
Valve
P
r
e
s
s
u
r
e
P1
P3
P2
0 1 2 3 4
Time
P
r
e
s
s
u
r
e

a
p
p
l
i
e
d

t
o

a
n
n
u
l
u
s
A
n
n
u
l
u
s

g
a
s

a
t

G
L
V
1
P
r
e
s
s
u
r
e

s
t
a
b
i
l
i
z
e
s

i
n

t
u
b
i
n
g
A
n
n
u
l
u
s

g
a
s

a
t

G
L
V
2
A
n
n
u
l
u
s

g
a
s

a
t

G
L
V
3
G
L
V
1

c
l
o
s
e
s
G
L
V
2

o
p
e
n
s
Annulus Pressure at GLV1 Depth
Tubing Pressure at GLV1 Depth
Annulus Pressure at GLV2 Depth
Tubing Pressure at GLV2 Depth
Annulus Pressure at GLV3 Depth
Tubing Pressure at GLV3 Depth
G
L
V
2

c
o
o
l
s
G
L
V
3

c
l
o
s
e
s
F
l
a
p
p
e
r

v
a
l
v
e

o
p
e
n
e
d
5 6 7 8
Fig. 11Schematic of the three-valve unloading process with the thermal valve. This process allows the thermal valve to be com-
patible with current well operations. Annulus and tubing pressures are plotted at each GLV depth over time during the unloading
process. The well sketch shows the condition at Time 0, with annulus and tubing full of kill uid. Annulus gas arriving at GLV2
(Time 3) has no effect in the tubing because the safety valve is closed.
82 February 2013 SPE Production & Operations
the ball-valve side extension and passes through a hole to the
bottom of the housing. The bottom end of the wire is xed to the
prototype housing by a bolt. Both attachment methods are recom-
mended by Dynalloy for attaching shape-memory alloy wires.
Tight seals are created at the entrance and exit of the ball valve
by using rubber O-rings. The ball valve sits in a cylindrical cavity
with O-rings on the top and bottom that deform to press tightly
against the ball valve.
Experimental Setup. The experimental setup is designed to
test the actuation temperature and hysteresis behavior of the pro-
totype valves and to act as a proof of concept of the thermally
actuated ball valve. Water is pumped from a storage tank, through
a water heater, through the prototype valve, and back into the
water-storage tank. The shape-memory alloy actuation wires are
heated by conduction through the metal housing. A thermocouple
senses the SMA wire temperature and a tilt sensor is mounted to
the ball-valve end extension to record the ball valve position. A
pressure transducer is mounted near the pump outlet to allow
closed-loop control of the pump speed.
The Labview software/hardware program is used to acquire
data from the sensors and supply necessary power for the sensors.
The centrifugal pump is supplied with a closed-loop PID control-
ler that relies on a pressure measurement at the pump outlet to
control the ow. The set pressure of the controller is manually
input as desired.
Experimental Results. Similar results were obtained from
tests of both prototypes, and results from the stainless-steel proto-
type are shown. Because a much larger temperature difference is
available downhole for cooling than heating, the prototypes were
only tested under heating. Fig. 13 shows an angular displacement
plot of the ball valve as a function of temperature for a typical ex-
perimental trial at realistic well temperatures. Data are ltered by
a second-order Butterworth low-pass lter. The plot shows that in
the 25

- to 85

C temperature range, the valve rotates closed 86

(from -6 to 80 on the plot). The dashed lines illustrate how, with

this specic SMA wire, a temperature change of 13

C is sufcient
to rotate the valve 66

as required to stop a backow event. The

valve rotates between 0

and 66

when heated from 58

to 71

C.
Conclusions
This paper introduced a new type of safety valve for use in gas-
lifted oil wells. Most ow-control safety valves currently used in
the offshore oil industry rely on active control or passive pres-
sure-induced actuation, but we showed that a thermally actuated
valve is a feasible option for many wells. We modeled steady-
state temperature proles in a gas-lifted oil well and veried
against experimental data. Results showed that a sufcient tem-
perature signal exists to actuate a shape-memory alloy safety
valve during an oil backow event. We described the design of an
SMA-actuated trunnion ball valve in the GLV assembly, and cre-
ated plastic and stainless-steel prototype valves. We tested valve
actuation at realistic shallow oil well temperatures and results
showed that, using Nitinol SMA wire, the safety valve is applica-
ble to wells with at least a 13

C gas/oil temperature difference.

This is a greater temperature difference than is available for most
wells as predicted by the thermal model, but other SMA wires
exist that require only a 3

C thermal signal to actuate. These

SMA wires would make the design feasible under the available
temperature signal predicted by the thermal model.
Nomenclature
a tubing wall thickness, m
b casing wall thickness, m
c
gas
gas-specic heat, J/(kg-K)
c
mix
mixture-specic heat, J/(kg-K)
c
oil
oil-specic heat, J/(kg-K)
h
ground
ground convection coefcient, W/(m
2
-K)
k
cem
casing conduction coefcient, W/(m-K)
k
gas
gas conduction coefcient, W/(m-K)
k
mix
mixture conduction coefcient, W/(m-K)
k
oil
oil conduction coefcient, W/(m-K)
k
tube
tubing wall conduction coefcient, W/(m-K)
_ m
gas
gas mass-ow rate, kg/s
_ m
mix
mixture mass-ow rate, kg/s
Nu Nusselt number, unitless
P
gas
gas pressure, Pa
_
Q
gasin
heat ow rate into gas control volume, W
_
Q
gasout
heat ow rate out of gas control volume, W
_
Q
ground
ground heat ow rate, W
_
Q
tube
tubing-to-annulus heat ow rate, W
R ideal gas constant, J/(mol-K)
R
case
casing radius, m
R
s
gas/oil volumetric ratio at injection depth, unitless
Rtotg thermal resistance across casing, K/W
Rtott thermal resistance across tubing, K/W
R
tube
tubing radius, m
T
gas
gas temperature, K
T
gassurf
surface gas temperature, K
T
mix
mixture temperature, K
T
res
reservoir temperature, K
T
tube
tubing temperature, K
1 cm
Top Section
Bottom
Section
Torsion
Spring
Trunnion
Ball Valve
SMA Wires
Fig. 12True-scale stainless steel prototype valve.
0 10 10 20 30 40 50 60 70 80
66
Ball Valve Angular Displacement (degrees)
30
40
50
60
70
80
G
L
V

T
e
m
p
e
r
a
t
u
r
e

(

C
)
1
3

C
Fig. 13Ball valve plot showing full closure and opening.
Dashed lines show the 13

C temperature change required for

the valve to rotate 66

.
February 2013 SPE Production & Operations 83
w well depth, m
X distance from injection depth, m
q
gas
gas density, kg/m
3
q
oil
oil density, kg/m
3
Acknowledgments
This work is supported by Chevron Corporation through the MIT-
Chevron University Partnership Program.
References
Ashby, M.F. 1996. Materials Selection in Mechanical Design. London:
Pergamon Press.
Atkins, P. and Jones, L. 2005. Chemical Principles: The Quest for Insight,
third edition. New York: W.H. Freeman and Company.
Bellarby, J. 2009. Well Completion Design, Vol. 56. Oxford, UK: Devel-
opments in Petroleum Science, Elsevier B.V.
Bible, H.V. 1977. Control system and improved pneumatically operated
temperature controlled valve construction therefor or the like. US Pat-
ent No. 4,016,853.
Brown, K.E. 1980. The Technology of Articial Lift Methods, Volume 2a:
Introduction of Articial Lift Systems, Beam Pumping: Design Analy-
sis Gas Lift. Tulsa, Oklahoma: Petroleum Publishing Company.
Carlsen, J.A., Stokka, S.., and Kleppa, E. 2010. Taking the Gas Lift
Valves to a New Level of Reliability. Paper OTC 20820 presented at
the Offshore Technology Conference, Houston, 36 May. http://
dx.doi.org/10.4043/20820-MS.
Das, D.K., Nerella, S., and Kulkarni, D. 2007. Thermal Properties of Pe-
troleum and Gas-to-liquid Products. Petroleum Science and Technol-
ogy 25 (4): 415425. http://dx.doi.org/10.1080/10916460500294556.
Deurig, T. 1990. Engineering Aspects of Shape Memory Alloys. Rushden,
UK: Reed Books Services.
Dynalloy, Inc. 2010. Dynalloy, Inc.: Makers of Dynamic Alloys, http://
www.dynalloy.com/.
Felzenbaum, A. 1992. Exponential model of the seasonal thermocline.
Phys. Oceanogr. 3 (1): 7579. http://dx.doi.org/10.1007/bf02198496.
Fridleifsson, I.B., Bertani, R., Huenges, E. et al. 2008. The possible role
and contribution of geothermal energy to the mitigation of climate
change. In IPCC Scoping Meeting on Renewable Energy Sources, Pro-
ceedings, Luebeck, Germany, 2025 January 2008, ed. O. Hohmeyer
and T. Trittin, 5980. Oxford, UK: International Geothermal Association
(IGA)/Elsevier. http://www.iea-gia.org/documents/FridleifssonetalI
PCCGeothermalpaper2008FinalRybach20May08_000.pdf.
Hart, W.B. and Williams, R.T. 1980. Temperature-responsive valve. US
Patent No. 4,227,646.
Holand, P. and Rausand, M. 1987. Reliability of Subsea BOP systems.
Reliability Engineering 19 (4): 263275. http://dx.doi.org/10.1016/
0143-8174(87)90058-8.
Homma, D. 1990. Valve driven by shape memory alloy. US Patent No.
4,973,024.
Kato, H., Yamamoto, T., and Hashimoto, S. 1999. High-Temperature Plas-
ticity of the b-phase in Nearly-Equiatomic Nickel-Titanium Alloys.
Mater. Trans. 40 (4): 343350.
Kelly, W.M. and Fredd, J.V. 1972. Fire safety valve. US Patent No.
3,659,624.
Lagoudas, D.C. ed. 2010. Shape Memory Alloys: Modeling and Engineer-
ing Applications. New York: Springer ScienceBusiness Media.
Ma, N., Hu, Z., Samuel, R. et al. 2006. Design and Performance Evalua-
tion of an Ultradeepwater Subsea Blowout Preventer Control System
Using Shape Memory Alloy Actuators. Paper SPE 101080 presented
at the SPE Annual Technical Conference and Exhibition, San Antonio,
Texas, USA, 2427 September. http://dx.doi.org/10.2118/101080-MS.
Mayeld, J.M. Jr. and Haselswerdt, V. 1982. Temperature controlled valve
mechanism and method. US Patent No. 4,356,833.
Place, D.E. 1979. Temperature responsive valve. US Patent No.
4,133,478.
Prensky, S. 2006. Recent Advances in LWD/MWD and Formation Evalu-
ation. World Oil 227 (March 2006).
Pymm, J. 1935. Temperature controlled valve. US Patent No. 2,004,636.
Sagar, R.K., Doty, D.R., and Schmidt, Z. 1991. Predicting Temperature
Proles in a Flowing Well. SPE Prod Eng 6 (6): 441448. SPE-19702-
PA. http://dx.doi.org/10.2118/19702-PA.
Sczerzenie, F. 2004. Consideration of the ASTM Standards for NiTi
Alloys. Proc., International Conference on Shape Memory and Supere-
lastic Technologies, Baden-Baden, Germany, 3-7 October, 203210.
Smith, P. and Zappe, R.W. 2004. Valve Selection Handbook: Engineering
Fundamentals for Selecting the Right Valve Design for Every Indus-
trial Flow Application, fth edition. Burlington, Massachusetts: Gulf
Professional Publishing.
Surbled, P., Clerc, C., Le Pioue, B. et al. 2001. Effect of the composition
and thermal annealing on the transformation temperatures of sputtered
TiNi shape memory alloy thin lms. Thin Solid Films 401 (12):
5259. http://dx.doi.org/10.1016/s0040-6090(01)01634-0.
Takacs, G. 2005. Gas Lift Manual. Tulsa, Oklahoma: PennWell Corporation.
Tanju, B.T. and Worman, P.J. 2009. Shape Memory Alloy Actuation. US
Patent Application No. 20090139727.
Tarvis, R.J. Jr. 1984. Temperature controlled valve. US Patent No.
4,454,983.
Eric Gilbertson holds BS and MS degrees in mechanical engi-
neering from the Massachusetts Institute of Technology, where
he is currently a PhD candidate. email: egilbert@mit.edu.
Franz Hover is currently Finmeccanica Associate Professor with
the MIT Department of Mechanical Engineering, with research
focusing on design methodologies for ocean engineering.
email: hover@mit.edu. He was a consultant to industry and
then a member of the research staff at MIT, where he worked
in fluid mechanics, biomimetics, and marine robotics. He holds
a BSME degree from Ohio Northern University and MS and ScD
degrees from the Joint Program in Applied Ocean Physics and
Engineering at the Woods Hole Oceanographic Institution
and the Massachusetts Institute of Technology.
Bryan Freeman is currently a production engineering advisor
at Hess Corporation. email: DFreeman@hess.com. He has
worked for 22 years in artificial lift in the areas of deepwater/
subsea in the US, west Africa, the Middle East, Australia, and
western Siberia. He holds BS and MS degrees in engineering
from the University of Texas at Austin.
84 February 2013 SPE Production & Operations