You are on page 1of 9

Optimisation of Three – Phase Separators: A Mathematical Model

Design
Ahmed, T., Russell, P. A and Hamad, F
School of Science and Engineering, Teesside University, Southfield Road Middleborough TS1 3BA United Kingdom

T.Ahmed@tees.ac.uk

Abstract: Multiphase separators are intrinsic to the recovery of crude oil. This work compares
extant separator design models with a newly developed procedure to optimise existing models
by minimising economic cost. The new method avoids the use of ‘fudge factors’ or undefined
variables in the solution of the design. A comparison between results obtained from this model
and four extant models is presented. The newly developed model gave results similar to models
based on retention time theory without the need for undertaking a laboratory testing.

Minimise: Cost (C)


(3)
1. INTRODUCTION = f(Capital cost, Operating cost)
The changing nature of the oil and gas industry dictates The capital cost is based from the work carried out by
that multiphase separators must be optimised to improve Powers (1990) for determining the capital cost of a
productivity and profitability. This can be achieved horizontal vessel. (see eq. 4).
through mathematical optimisation by selecting the best 2
𝐶 = 𝑓 ∗ 𝐹𝑐 ∗ 𝜌𝑠 ∗ (𝜋 ∗ 𝐷𝑚 ∗ 𝑉𝐿 + 2 ∗ 𝐹𝑎
element from a set of available elements. Mathematical 3 (4)
∗ 𝐹ℎ ∗ 𝐷𝑚 )
Optimisation has been used extensively in chemical and
process design. However, to achieve this, the problem We can break down eq. 4 into three different types of
must first be designed to fit into the following general variables;
form;
2.1.1 Input variables
𝑂𝑝𝑡𝑖𝑚𝑖𝑧𝑒: 𝑦 = 𝑓(𝑋1 , 𝑋2 … , 𝑋𝑛 ) (1)
≤ These variables are divided into calculated input
𝑆𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜: 𝑔𝑗 (𝑥1 , 𝑥2 , … , 𝑥𝑛 ) {=} 𝑏𝑗 𝑗 variable(s), manufacturer supplied input variables and
(2) constants. The vessel shell thickness is a calculated input

= 1,2, … 𝑚 variable. Vessel shell can be either thin or thick walled
Equation 1 is the objective function which is and the rule of thumb used to differentiate between the
maximised or minimised subject to Eq. 2, which is the set two categories is that, in the former, the vessel’s diameter
of constraints imposed on the solution. The variables x1, is at least ten time the vessel thickness, otherwise, it is
x2, …, xn are the set of decision variables. The constraints categorised as thick-walled vessel. Most separator
are expressed as equalities and inequalities see Eq. 2. vessels in the process industries are thin – walled. The
Satisfying all the constraints renders a feasible solution. design, fabrication and testing of pressure vessels is
commonly based on the American Society of Mechanical
Winston (1994) reported that different algorithms can Engineers (ASME) code for the design of pressure
be used to solve Mathematical optimisation. Therefore, it vessels. This code is widely used in the oil and gas
is important to choose the right algorithms for each industry in the sizing of three-phase separators among
optimisation problem. Therefore, due to the nonlinearity other pressure vessels. Hence vessel shell thickness (thin
in both the objective function and the constraints, the walled) can be calculated using eq. 5 (Silla, 2003)
Generalised Reduced Gradient (GRG) was chosen to
𝑃𝐷 ∗ 𝐷𝑖
optimize the separator design. 𝑓= + 𝑡𝑐 (5)
2𝜎𝐸 − 1.2𝑃𝐷
2. MODEL THEORETICAL DESIGN The design pressure (PD) is typically the maximum of
This section describes the theoretical design of the either the operating pressure plus 10% or the operating
objective function and the constraints used in pressure plus 15 to 30 psi. Joint efficiency(E) ranges from
determining the optimum separator size. 0.6 to 1, with 1 for 100% x – rayed joint. The tensile
strength (𝜎) of carbon steel commonly used due to its
2.1 Objective Function ability to withstand high temperature is 950bar.
It was decided to base this model on an objective Corrosion allowance (tc) ranges from 1.6mm to 3.2mm
function formed from capital and operating cost variables 𝑃𝐷 = max(1.1 ∗ 𝑃, 𝑃 + 200,000) (6)
(see eq. 3). Initially, the operating cost was set to zero
because other models concentrate on the capital cost only. The second group of input variables are supplied by the
vessel manufacturer. These include the cost per unit mass
to manufacture the vessel (Fc), factor for determining
surface area of head from the vessel diameter (Fa) and the

1
Optimisation of Three-Phase Separators: A Mathematical Model Design

ratio of cost per unit mass to manufacture vessel head 𝐿𝑔𝑟


𝑡𝑟 = ⁄𝑈 (10)
compared with that of the vessel shell (Fh). Default 𝑔
values were included in the model based on Grodal and 𝑄𝑔
Realff (1999). 𝑈𝑔 = ⁄𝐴 (11)
𝑔
Finally, the last group of inputs in the objective Retention time is equal to the length of the gravity
function are the constants. These are pi (π) and the density settling section divide by the flow velocity as in eq. 10.
of the steel, default is carbon steel. We can calculate the flow velocity by dividing the
2.1.2 Fitted variable flowrate by the area as in eq. 11. Substituting eq. 12 and
13 into 9 and solving for Lgr gives eq. 14 which is the
Vessel mean diameter is dependent on the vessel length required for gravity settling of the liquid droplets
internal diameter. out of a gas continuous phase.
𝐷𝑖 + (𝐷𝑖 + 2𝑓) 𝐿𝑔𝑟 𝐴𝑔
𝐷𝑚 = (7) 𝑡𝑟 = ⁄𝑄 (12)
2 𝑔

The internal diameter (Di) is fitted by the GRG non- (𝐷𝑖 − ℎ𝑁𝑂𝐿 )
linear solver in excel subject to constraints. 𝑡𝑜𝑔 = (13)
𝑈𝑜𝑔
2.1.3 Calculated variable
𝑄𝑔 (𝐷𝑖 − ℎ𝑁𝑂𝐿 )
The main calculated variable in the cost equation is the 𝐿𝑔𝑟1 = (14)
𝑈𝑜𝑔 (VA − 𝐴𝑁𝑂𝐿 )
separator length. The separator length is the sum of the
lengths of the separator inlet, gravity settling section and The settling velocity (in this case, of oil droplets out of
the outlet sections. The gravity settling section length is gas phase) is determined by equating the gravity force to
calculated from the eq. 9 – 19. It is important to point out the drag force. This calculation is an iterative process
that the design of the separator inlet is not considered which start with an assumption for the initial value of the
here, however, it is believed that whatever the inlet device drag coefficient. The terminal velocity is calculated and
chosen, the length should not exceed 1m and is set as used to calculate the Reynolds number which in turn is
default. Vessel length can therefore be calculated from used to calculate the drag coefficient. This value is then
equation 8 below; used as the input into the terminal velocity equation and
the procedure is repeated until the difference between the
𝑉𝐿 = 𝐿𝑖 + 𝐿𝑔𝑟𝑣 + 𝐿𝑜 (8) calculated value equals the assumed value. Terminal
Length of the gravity settling section is obtained as the velocity, Reynolds number and drag coefficient are
maximum length required for a dispersed phase to settle calculated using eq. 15 – 17.
out of the continuous phase in the three – phase separator.
This length can be calculated by setting the phase 1.333 ∗ 𝑑𝑝 ∗ 𝑔 ∗ (𝜌𝑜 − 𝜌𝑔
𝑢 𝑇 = √[ ] (15)
retention time in the vessel equal to the bubble / droplet 𝐶𝐷 ∗ 𝜌𝑔
rising / settling time (see eq. 9). For this model, rising /
settling paths were assumed and used to calculate the 24 3
𝐶𝐷 = 0.34 + + (16)
lengths required for separation of the gas, oil and water. 𝑁𝑅𝑒 √𝑁𝑅𝑒
In a continuous gas phase, oil and water settle, water 𝜌 ∗ 𝑈 ∗ 𝑑𝑝
settle faster than oil so the oil droplet settling is 𝑁𝑅𝑒 = (17)
µ
controlling and is considered from the vessel top to the
Normal Operating Level.
In a continuous oil phase, the gas bubbles rise and the The length required for settling water droplets from the
water droplet settle. The gas bubble rise faster so the oil continuous phase and that of rising oil droplets from
water droplet settling is controlling and is considered water continuous phase are obtained using equations 18
from the Normal Operating Level to the Normal Interface and 19 respectively;
Level. 𝑄𝑜 (ℎ𝑁𝑂𝐿 − ℎ𝑁𝐼𝐿 )
In a continuous water phase, the gas bubble and the 𝐿𝑔𝑟2 = (18)
𝑈𝑤𝑜 (𝐴𝑁𝑂𝐿 − 𝐴𝑁𝐼𝐿 )
liquid droplet rise. The gas bubble rise faster and so the
oil droplet rising from the bottom of the vessel to the 𝑄𝑤 ℎ𝑁𝐼𝐿
𝐿𝑔𝑟3 = (19)
Normal Interface Level is controlling and therefore 𝑈𝑜𝑤 𝐴𝑁𝐼𝐿
considered. Length of the separator outlet section is calculated as;
Equations 9 to 14 below are examples developed based 𝐿𝑂 = 2(𝑑𝑛𝑜 + 𝑑𝑛𝑤 ) + 𝐿𝑤𝑒𝑖𝑟
on the separation of liquid droplet from the gas
continuous phase. The same procedure is used to The oil and water outlet nozzles can be calculated
determine the length required for gravity settling for the using Monnery and Svrcek, (1994) formula for
oil and water phases respectively. calculating nozzle sizes (see eq. 21);
𝑡𝑟 ≥ 𝑡𝑜𝑔 (9) 𝑑𝑛𝑝 = 0.146√𝑄 ∗ 𝜌0.5 (21)

2
Optimisation of Three-Phase Separators: A Mathematical Model Design

Once the objective function is calculated, the next step is the outlet nozzle, the extractor’s efficiency will be
to determine the constraints. reduced since most of the flow will be directed towards
the centre.
2.2 Constraints
The constraints set to the objective function are We can obtain the height of high high liquid level from
grouped into three; the area of High High Liquid Level. Powers (1990) and
Svrcek and Monnery (1993) are among the several works
2.2.1 Phase Outlet Constraint that reported on how to calculate dimensionless chord
The first group of constraints are set to ensure the height from dimensionless chord area and vice versa. In
various phases leave the separator through their this model, two user defined functions referred to as
designated outlets. Firstly, to avoid oil from escaping “HTOA” and “ATOH” were developed. These functions
through the gas outlet, there should be a minimum length are explained in the excel model development sections of
between the high high liquid level (hHHLL) and the inlet this paper.
of the mist extractor. Secondly, to avoid gas leaving the 𝐴𝐻𝐻𝐿𝐿
separator from the liquid outlets, there should be a 𝐻𝐻𝐻𝐿𝐿 = 𝐴𝑇𝑂𝐻 ( ) ∗ 𝐻𝑇 (22)
𝑉𝐴
distance between the weir height (hWH) and the low low To calculate hHHLL above we need to calculate
liquid level (hLLLL). Thirdly, to avoid water level from AHHLL. This can be calculated from the Norsok
reaching and overflowing the weir and leaving through Standard that states “In sizing the separators, the
the oil outlet, there should be a distance between the weir equivalent residence time between normal and alarm
height (hWH) and the high high interface level (hHHIL). level and between alarm and trip level should not be less
Finally, to avoid oil from leaving through the water than 30 seconds or 80mm for both high and low ranges”.
outlet, there should be a distance between the low low
interface level (LLIL) and the bottom of the vessel. (𝑄𝑜 + 𝑄𝑤 )∆𝑡𝑁𝑆 ≤ 𝑉𝐿 ∗ (𝐴𝑎 − 𝐴𝑏 ) (23)

a = (HHLL, HLL, NOL, LLL)

b = (HLL, NOL, LLL, LLLL)


Substituting a and b for HHLL and HLL into eq. 36
and solving for AHHLL gives eqn. 37
(𝑄 + 𝑄𝑜 )∆𝑡𝑁𝑜𝑟
𝐴𝐻𝐻𝐿𝐿 = ( 𝑜 ) + 𝐴𝐻𝐿𝐿 (24)
𝑉𝐿
AHLL can calculated using the Norsok Standard
equation mentioned above or the Holdup and Surge
volume. The maximum value from the two equations is
chosen for the area of high liquid level. Holdup volume
is the volume from the vessel normal operating level to
the low liquid level while Surge Volume is the volume
from the vessel normal operating level to the high liquid
level. Holdup and surge volume can be used to set
distance from the Normal operating level to the Low and
Figure 1: Separator Outlet Section showing outlet constraints
High liquid levels respectively. The Holdup and surge
volumes are applicable to interface levels too. We can
These constraints can be defined as;
define the holdup and surge volumes for normal liquid
hMEI – hHHLL ≥ 0.10m
levels as;
hLLLL – hWH ≥ 0.10m
𝑉𝐻𝑜𝑙𝑑𝑢𝑝 = 𝑉𝐿 ∗ (𝐴𝑁𝑂𝐿 − 𝐴𝐿𝐿𝐿 ) (25)
hWH – hHHIL ≥ 0.10m
hLLIL ≥ 0.10m 𝑉𝑠𝑢𝑟𝑔𝑒 = 𝑉𝐿 ∗ (𝐴𝐻𝐿𝐿 − 𝐴𝑁𝑂𝐿 ) (26)

1. Avoid liquid from leaving through the gas Therefore, we can calculate Area of HLL as;
outlet (hMEI – hHHLL ≥ 0.10m) (𝑄𝑜 + 𝑄𝑤 )∆𝑡𝑁𝑜𝑟
𝐴𝐻𝐿𝐿 = + 𝐴𝑁𝑂𝐿 (27)
The difference between the height of mist extractor 𝑉𝐿
inlet and the height of the high high liquid level should 𝑉𝑠𝑢𝑟𝑔𝑒
be more than the safety factor which is 0.10m. Different 𝐴𝐻𝐿𝐿 = + 𝐴𝑁𝑂𝐿 (28)
𝑉𝐿
literature suggested different safety factors between the
mist extractor and the liquid level. Here, an average of To calculate the AHLL, the ANOL need to be
0.10m is used for this model. calculated. This can be calculated from the hNOL using
the dimensionless chord height formula as;
The Mist Extractor Inlet is obtained by subtracting the
vessel diameter by 0.3m. The mist extractor height is ℎ𝑁𝑂𝐿
𝐴𝑁𝑂𝐿 = 𝐻𝑇𝑂𝐴 ( )
usually set to be 0.15m (6inches). Another 0.15m is 𝐻𝑇 (29)
allowed from the top of the vessel to the mist extractor ∗ 𝐴𝑇
outlet. This height is set to obtain a uniform flow
distribution through the extractor. If placed too close to

3
Optimisation of Three-Phase Separators: A Mathematical Model Design

𝜋 ∗ 𝐷2 𝑄𝑤
𝐴𝑇 = (30) 𝑉𝐻𝑜𝑙𝑑𝑢𝑝 ( ) ≤ 𝑉𝑙 ∗ (𝐴𝑁𝐼𝐿 − 𝐴𝐿𝐼𝐿 ) (39)
4 𝑄𝑜
Di and hNOL are assumed which are then set by the 𝑄𝑤
excel solver. 𝑉𝑆𝑢𝑟𝑔𝑒 ( ) ≤ 𝑉𝑙 ∗ (𝐴𝐻𝐼𝐿 − 𝐴𝑁𝐼𝐿 ) (40)
𝑄𝑜
2. Avoid gas from leaving through the liquid outlets Rearranging eqn. 40 gives;
(hLLLL – hWH ≥ 0.10m) 𝑉𝑆𝑢𝑟𝑔𝑒 ∗ 𝑄𝑤
𝐴𝐻𝐼𝐿 =
The difference between the low low liquid level and 𝑉𝑙 ∗ 𝑄𝑜 (41)
the weir height is set to be greater than 0.10 m to avoid + 𝐴𝑁𝐼𝐿
the gas phase from leaving through any of the liquid ANIL can be calculated from the hNIL
outlets. Once the height of the low low liquid level is ℎ𝑁𝐼𝐿
calculated, the weir height can then be set as the low low 𝐴𝑁𝐼𝐿 = 𝐻𝑇𝑂𝐴 ( ) ∗ 𝐴𝑇 (42)
liquid level minus 0.10m. Low low liquid height can be 𝐻𝑇
calculated from the area of the low low liquid level which hNIL is assumed from the initial guesses.
can be calculated by rearranging equation 23.
4. Avoid Oil from leaving through the Water outlet
(𝑄𝑂 + 𝑄𝑊 )∆𝑡𝑁𝑆 (hLLIL ≥ 0.10m)
𝐴𝐿𝐿𝐿𝐿 = 𝐴𝐿𝐿𝐿 − ( ) (31)
Oil being lighter than water rise to the top of the water
𝑉𝐿
continuous phase. To avoid the oil from leaving through
(𝑄𝑂 + 𝑄𝑊 )∆𝑡𝑁𝑆 the water outlet, the height of the low low interface level
𝐴𝐿𝐿𝐿 = 𝐴𝑁𝑂𝐿 − ( ) (32)
𝑉𝐿 should be at least 0.10m. The height of the low low
interface level can be calculated as;
From eq. 25; 𝐴𝐿𝐿𝐼𝐿
𝑉𝐻𝑜𝑙𝑑𝑢𝑝 ℎ𝐿𝐿𝐼𝐿 = 𝐴𝑇𝑂𝐻 ( ) ∗ 𝐻𝑇 (43)
𝐴𝐿𝐿𝐿 = 𝐴𝑁𝑂𝐿 − (33) 𝐴𝑇
𝑉𝐿
ANOL is calculated from above 𝑄 ∗ ∆𝑡𝑁𝑆
𝐴𝐿𝐿𝐼𝐿 = 𝐴𝐿𝐼𝐿 − ( 𝑤 ) (44)
ℎ𝑊𝐻 = ℎ𝐿𝐿𝐿𝐿 − 0.1𝑚 (34) 𝑉𝑙
Water nozzle diameter (dnw) has been calculated using
3. Avoid Water from leaving through the Oil outlet equation 34
(hWH – hHHIL ≥ 0.10m) 𝑉𝐻𝑜𝑙𝑑𝑢𝑝 ∗ 𝑄𝑤
𝐴𝐿𝐼𝐿 = 𝐴𝑁𝐼𝐿 − (45)
In vessels with liquid outlets at the bottom of the 𝑄𝑜 ∗ 𝑉𝑙
vessel, weir is used to prevent the water phase from 𝑄 ∗ ∆𝑡𝑁𝑜𝑟
leaving the vessel through the oil outlet. To ensure this, 𝐴𝐿𝐼𝐿 = 𝐴𝑁𝐼𝐿 − ( 𝑤 ) (46)
the difference between the weir height and the high high 𝑉𝑙
interface level is set to be equal to or greater than 0.10m. ℎ𝑁𝐼𝐿
𝐴𝑁𝐼𝐿 = 𝐻𝑇𝑂𝐴 ( ) ∗ 𝐴𝑇 (47)
Failure to enforce this constraint in the separator might 𝑉𝐻
lead to the water flowing over the weir and escaping 2.2.2 Transportation Constraints
through the oil outlet at certain conditions. Therefore,
hHHIL is calculated from AHHIL using; The second group of constraints are new and are
default values. They were derived from the maximum
𝐴𝐻𝐻𝐼𝐿 dimensions allowed for road transport in the UK and US.
ℎ𝐻𝐻𝐼𝐿 = 𝐴𝑇𝑂𝐻 ( )
𝐴𝑇 (35) Most of the work carried out on the sizing of three-phase
∗ 𝐻𝑇 separators does not consider the transportation of these
vessels from the manufacturing to the operation sites.
As mentioned earlier, Norsok guidelines are applicable
Hence, it is necessary to ensure all manufactured vessels
to interface levels. Hence eq. 36 can be represented for
are within the road transport limits. From the literature,
interface levels as;
the maximum length that can be transported by road is
𝑄𝑤 ∆𝑡𝑁𝑆 ≤ 𝑉𝑙 ∗ (𝐴𝑥 − 𝐴𝑦 ) (36) 18.75m and a diameter of 4.23m (Loh, Lyons and White,
Where; x = (HHIL, HIL, NIL, LIL) 2002; UK Government Revenue and Customs; US
y = (HIL, NIL, LIL, LLIL) Department of Transport)
Vl = (𝐿𝐼 + 𝐿𝑔𝑟 + 2𝑑𝑛𝑤 ). Therefore, the following constraints were developed;
AHHIL can be calculated by substituting x and y for The maximum separator diameter (VD) should not be
HHIL and HIL into eq. 36 which becomes; more than 4.23m
𝑄 ∗ ∆𝑡𝑁𝑆
𝐴𝐻𝐻𝐼𝐿 = 𝑤 + 𝐴𝐻𝐼𝐿 (37) VD = Di + 2tc (48)
𝑉𝑙
AHIL can be calculated using the maximum of eqns. 51 The maximum separator length (LT) should not be
and 54; more than 18.75m
𝑄 ∗ ∆𝑡𝑁𝑆 LT = VL + 2(HL + tc) (49)
𝐴𝐻𝐼𝐿 = 𝑤 + 𝐴𝑁𝐼𝐿 (38)
𝑉𝑙 HL = Di/4 (50)
AHIL can also be calculated using the Holdup and surge 2.2.3 Positive Decision Variables Constraints
volume. For interface levels, eqns. 25 and 26 can be
represented as; The third group of constraints are the decision

4
Optimisation of Three-Phase Separators: A Mathematical Model Design

variables. For optimisation to proceed, it is necessary to


set some inputs at the start and to constraint them towards
a possible solution. Therefore, the set inputs i.e. internal
diameter (Di), length of gravity settling section (Lgr),
height of normal operating level (hNOL) and height of
normal interface level (hNIL) are constrained to be
greater than 0.
The theoretical model developed was built into excel
spreadsheet to obtain numerical results. The following
section explains how the excel spreadsheet was
developed.

3. EXCEL SPREADSHEET DESIGN


The excel spreadsheet design is grouped into three Figure 2: Process Flow Diagram for Ahmed et al., (2016) Model
different parts;
Optimisation Solver: Different solvers are available in 4. COMPARISON WITH OTHER MODELS
excel for optimisation problems. In this model, the To determine if the developed model (referred to as
Generalised Reduced Gradient (GRG) solver was chosen Ahmed et al. 2016) was producing representative results
because of the non-linearity in the separator design. GRG it was decided to compare it with 4 extant models. The
solver uses iterative numerical methods that involve models considered were; Abdel-Aal and Aggour (2003),
using trial values for the adjustable cell and observing the Arnold and Stewart (1998), Svrcek and Monnery (1994),
results calculated by the constraint cells and the optimum and Dokianos (2015). An excel spreadsheet was built for
cell. Each of this trial is known as an iteration. Because each method and used to produce simulations for a fixed
the developed model consists of many adjustable cells set of conditions.
and constraints, ordinary trial and error will take
extremely long time, GRG solver performs an extensive 4.1 Comparison at Fixed Gas and Water Flows
analysis of the observed output and their rate of change Initially it was decided to fix the gas and water
as the inputs changes, to guide the selection of trial flowrates at 5886m3/hr and 33.12m3/hr respectively and
values. vary the oil flowrate from 19.8 to 46.2 m3/hr. to
investigate the effect of oil fraction on the separator size.
User Defined Functions (UDF): Two UDF’s were From Figure 3, Abdel Aal and Aggour (2003) and Arnold
developed into excel visual basic for applications to and Stewart (1998) produce very similar results. A linear
speed-up the calculation processes. These are; relationship is seen between the increase in oil flow rate
and the length of the vessel. This is not surprising because
i. Dimensionless Chord Height: Calculation both models were developed to use droplet settling theory
of dimensionless chord height from the for gas–liquid separation and retention time theory for
dimensionless chord area (ATOH) is an liquid–liquid separation. In both cases, it was found that
iterative operation which is converging liquid capacity constrained the design. Given this the
independently to the initial value of φ (See diameter is fixed and integral to both models is the fact
eqn. 51). that length is set as being proportional to the product of
2πA∗ − φi + sin φi flowrate and retention time.
φi+1 = φi − (51)
cosφi − 1 Dokianos (2015) model produces separator
dimensions that do not change with increasing oil
ii. Dimensionless Chord Area: A function for flowrate. This method is more closely constrained than
calculating the dimensionless chord area the others given the length to diameter ratio has to be four,
from the dimensionless chord height the separator has to operate at 80% full of liquids and the
referred to as “HTOA”. liquid phase in the separator is assumed to operate under
2h∗ laminar conditions. Taken together these constraints fix
φ = 2 cos −1 (1 − ) (52) the separator dimensions for all oil flows.
D
(φ − sin φ) Finally, the Ahmed et al., (2016) and Svrcek and
A∗ = (53)
2π Monnery (1994) models were developed from droplet
settling theory for both gas–liquid separation and liquid–
The two functions were developed to easily calculate liquid separation. It is not therefore surprising that they
the dimensionless chord area from the dimensionless produce similarly shaped traces. What is interesting is
chord height and vice versa without the trouble of using that while the Ahmed et al (2016) data gives a similar
more complicated ways as reported by Power (1990) and diameter to the retention time-based methods, the Svrcek
Svrcek and Monnery (1993). Figure 2 presents the flow and Monnery (1994) results give a diameter which is
diagram for the developed procedure. systematically larger than the other methods. A further
graph (fig 4) was plotted where L/D is plotted as a
function of oil flow rate. In this case the trend obtained
from Ahmed et al., (2016) more closely matches that

5
Optimisation of Three-Phase Separators: A Mathematical Model Design

obtained from the retention time models within the as low ratios lead to plug flow causing poor separation.
expected range of 3. The results of Svrcek and Monnery The Ahmed et al (2016) results show similar trends for
(1994) are systematically low and show a slight decrease the retention time methods. The length of separators
in the ratio as flow is increased. The consequences of modelled using Abdel Aal and Aggour (2003) and Arnold
using the Svrcek and Monnery (1994) method are that the and Stewart (1998) increases linearly as the water
increase in diameter would produce a more expensive flowrate increases. Ahmed et al., (2016) model produces
separator because of the increase in material thickness separator length that also increase as the water flowrate
required to maintain the internal pressure. It may increases. Dokianos (2015) was found to produce
therefore be concluded that the use of optimisation of cost constant dimensions as in the previous group. All the
in the Ahmed et al., (2016) model leads to a design which diameters were found to be within 1.5 to 1.7 meters
is more in line with those predicted by retention theory excluding Svrcek and Monnery (1994). This same reason
without the need for experimentation. makes its length to be smaller than those of the other
models. Slenderness ratio was found to be within the
4.2 Comparison at Fixed Gas and Oil Flows
range of 1.5 to 5 with Svrcek and Monnery (1994) having
The next step was to fix the gas and oil flowrates at the lowest.
5886m3/hr and 46.2m3/hr respectively and vary the water
flowrate from 6.6 to 33.12 m3/hr. to investigate the effect 4.3 Comparison at Fixed Oil and Water Flows
of water fraction on the separator size. These results are
Finally, it was decided to keep the oil and water flows
reported in Figure 5 and Figure 6. Again, the diameters
constant and vary the gas flow to check the effects of gas
predicted by Svrcek and Monnery (1994) method are
flows on the separator sizes. Results obtained in figure 7
systematically higher than the other methods. When
and 8 indicated that liquid capacity constraint the design
considering the slenderness ratio, the Svrcek and
in all the three models. Hence all the models produce
Monnery (1994) trend is again systematically low and
constant dimensions. Slenderness ratio was found to be
does not show a rise with increasing flow rate. The lower
between 2 to 5. With Dokianos (2015) at 4 and Svrcek
slenderness ratios of less than 2 across the range predicted
and Monnery (1994) at 2.
Svrcek and Monnery (1994) could be a cause for concern

Figure 3: A graph of length and diameter against Oil flowrate

AA – Abdel Aal and Aggour (2003); Arnold – Arnold and Stewart (2008); Svrcek – Svrcek and Monnery (1994)
Ahmed – Developed model; Williams – Dokianos (2015)

6
Optimisation of Three-Phase Separators: A Mathematical Model Design

Figure 4: A graph of slenderness ratio against Oil flowrate

10
AA (D) AA (L) Arnold (D)
Arnold (L) Svreck (D) Svreck (L)
Length and Diameter (m)

Ahmed (D) Ahmed (L) Williams (D)


8 Willaims (L)

0
0 10 20 30 40
Water flowrate(m3/hr)
Figure 5: Diameter and Length against water flowrate

5.0

4.0

3.0
L/D

2.0

AA (L/D) Arnold (L/D) Svreck (L/D)


1.0
Ahmed (L/D) Williams (L/D)

0.0
0 5 10 15 20 25 30 35
Water flowrate (m3/hr)
Figure 6: Slenderness ratio against Water flowrate

7
Optimisation of Three-Phase Separators: A Mathematical Model Design

10
AA (D)

Length and Diameter (m)


AA (L)
8
Arnold
(D)
6 Arnold (L)

Svreck (D)
4
Svreck (L)

Tariq (D)
2

0
0 2000 4000 6000 8000 10000
Gas Flowrate (m3/hr)
Figure 7: Diameter and Length against gas flowrate

5.0

4.0

3.0
L/D

2.0

AA (L/D) Arnold (L/D)


1.0 Svreck (L/D) Tariq (L/D)
Williams (L/D)

0.0
0 2000 4000 6000 8000 10000
Gas Flowrate (m3/hr)
Figure 8: Slenderness ratio against Gas flowrate

CONCLUSION 2. Svrcek and Monnery (1993): Similar to Arnold and


Stewart and Abdel Aal and Aggour, this model is
In conclusion, a new mathematical model was designed also based on the assumption of half-filled
that was aimed at optimising three-phase separator separators. Separators sized using Svrcek and
design. The model was based on droplet settling theory Monnery models will tend to be more-costly due to
using Generalised Reduced Gradient (GRG) Nonlinear its wider diameters obtained in all the three sets of
optimisation solver in excel. Similarly, a comparative simulations conducted. The model is also based on
study was carried out with four other extant models. a number of tables look-ups and manual iterations
Based on the comparison conducted in this work, the which makes it less accurate.
following conclusions can be made; 3. Ahmed et al., (2016) model gives more details such
1. Arnold and Stewart (1998) and Abdel-Aal and as the nozzle sizes, liquid levels, settling velocities
Aggour (2003): These two models are based on the of Oil and Water, weir height and vessel wall
assumption that the separator operates half full. thickness. The newly developed model found out
They give very similar results which are directly that operating the separator at a liquid level of three
proportional to the retention time and liquid flow – quarter from the bottom of the vessel is more
rate except with high gas flows. Equal changes in beneficial rather than the mid-level. However, when
lengths is observed for every change in liquid flow compared with other models at high gas ratios,
rate. No change was found when the gas flow rate results obtained were inconclusive and hence need
increase for the chosen flowrates. The models do more work.
not give adequate information about the separator.

8
Optimisation of Three-Phase Separators: A Mathematical Model Design

REFERENCES Manning, F.S. & Thompson, R.E. 1995, Oilfield processing of


petroleum: Crude oil, Pennwell books, Oklahoma.
Abdel-Aal, H.K., Aggour, M.A. & Fahim, M.A.
2015, Petroleum and gas field processing, CRC Press. Monnery, W.D. & Svrcek, W.Y. 1994, "Successfully specify
three-phase separators", Chemical Engineering
Arnold, K.E. & Koszela, P.J. 1987, Droplet Settling vs. Progress, vol. 90, pp. 29-29.
Retention Time Theories for Sizing Oil/Water Separator,
Offshore Technology Conference, Houston, Texas. Norwegian Technology Centre 2001, Process Systems,
Norwegian Technology Centre, Oslo, Norway.
Arnold, K. & Stewart, M. 2008, "Chapter 5 - Three-Phase Oil
and Water Separation" in Surface Production Operations Powers, M.L. 1990, "Analysis of gravity separation in
(Third Edition), eds. K. Arnold & M. Stewart, Gulf freewater knockouts", SPE Production Engineering, vol.
Professional Publishing, Burlington, pp. 244-315. 5, no. 01, pp. 52-58.

Arnold, K.E. & Koszola, P.J. 1990, "Droplet-Settling vs. Sayda, A.F. & Taylor, J.H. 2007, "Modeling and Control of
Retention-Time Theories for Sizing Oil/Water Separator", Three-Phase Gravilty Separators in Oil Production
vol. 5, no. 01, pp. 59-64. Facilities", 2007 American Control ConferenceIEEE, , pp.
4847.
Boiler, A. & Code, P.V. 1992, "Section VIII-Division
I", Pressure Vessels, Section UG-27, pp. 24. Seider, W.D., Seader, J., Lewin, D.R. & Widagdo, S. 2008,
"Product and Process Design Principles: Synthesis,
Dokianos, W.P., 2015, March. A simplified approach to sizing Analysis, and Design, Section 9.3".
2 and 3 phase separators for low GOR and low pressure
onshore production batteries. In SPE Production and Silla, H. 2003, Chemical process engineering: design and
Operations Symposium. Society of Petroleum Engineers. economics, CRC Press, United States of America.

Grødal, E.O. & Realff, M.J. 1999, "Optimal design of two-and Smith, H.V. 1987, "Oil and Gas Separators (1987 PEH
three-phase separators: A mathematical programming Chapter 12)" in Society of Petroleum Engineers.
formulation", SPE Annual Technical Conference and
ExhibitionSociety of Petroleum Engineers. Soffian, R.M. & Niven, T. 1993, "Emulsion Treatment
Program", SPE Asia Pacific Oil and Gas
Kaasa, O. 1995, SYNENERGI - Production Chemicals and ConferenceSociety of Petroleum Engineers.
Process Design, Society of Petroleum Engineers.
Svrcek, W. & Monnery, W. 1993, "Design Two-Phase
Loh, H., Lyons, J. & White, C. 2002, Process equipment cost Separators Within", Chemical Engineering Progress.
estimation.Final Report, National Energy Technology
Center, . Winston, W.L. 1994, "Operations research applications and
algorithms", Inc., Belmont, CA.

You might also like