QUESTION 1

CONTROL OF MULTI-PHASE FLOW IN OFFSHORE OIL & GAS PIPELINES

2
question 1 – Process information

Consider a offshore upstream oil & gas pipeline process. Here, undesired cyclic pressure and flow oscillations (referred to as ”severe
slug flow”) can damage the process. These oscillations can be eliminated by controlling a topside choke valve, z. However, several
measurements are available, and it is uncertain which sensor type and location is the most optimal for anti-slug control. The
possibilities are:
1. Topside pressure measurement (𝑃𝑡 ): Cheap and minimum maintenance
2. Topside mass flow measurement (𝜔𝑡 ): Expensive and low maintenance
3. Lowpoint (also called bottom) pressure measurement (𝑃𝑏 ): Very expensive and high maintenance
• The figure below shows the multi-phase pipeline process with the location of the three different possible measurements
(process outputs): 𝑃𝑏 , 𝑃𝑡 , 𝜔𝑡 , as well as the topside choke valve, z, (process input). As an example of a possible control
configuration, the lowpoint pressure signal, 𝑃𝑏 , is fed into the controller and then into the actuator (z).

3
question 1 – Control of multi-phase flow in offshore oil & gas pipelines

The process is non-linear, but by linearizing the system at the average flow rates and choke valve openings at 30% (z=0.3), then we
can obtain three different transfer functions:
𝑃𝑡 (𝑠) −15.22𝑠 3 −19.01𝑠 2 −0.9666𝑠−0.366
1. 𝐻1 𝑠 = 𝑇𝑜𝑝𝑠𝑖𝑑𝑒 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 𝑏𝑎𝑟 / 𝑐ℎ𝑜𝑘𝑒 𝑜𝑝𝑒𝑛𝑖𝑛𝑔 − =
𝑧(𝑠)
=
𝑠 4 +12.29𝑠 3 +20.26𝑠 2 −0.1159𝑠+0.149
𝑘𝑔 𝜔𝑡 (𝑠) 324𝑠 4 +1977𝑠 3 +4729𝑠 2 +12.11𝑠
2. 𝐻2 𝑠 = 𝑇𝑜𝑝𝑠𝑖𝑑𝑒 𝑚𝑎𝑠𝑠 𝑓𝑙𝑜𝑤
𝑠
/ 𝑐ℎ𝑜𝑘𝑒 𝑜𝑝𝑒𝑛𝑖𝑛𝑔 − =
𝑧(𝑠)
=
𝑠 4 +12.29𝑠 3 +20.26𝑠 2 −0.1159𝑠+0.149
𝑃𝑏 (𝑠) −15.91𝑠 3 −20.14𝑠 2 −8.999𝑠−0.3909
3. 𝐻3 𝑠 = 𝐿𝑜𝑤𝑝𝑜𝑖𝑛𝑡 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 𝑏𝑎𝑟 / 𝑐ℎ𝑜𝑘𝑒 𝑜𝑝𝑒𝑛𝑖𝑛𝑔 − =
𝑧(𝑠)
=
𝑠 4 +12.29𝑠 3 +20.26𝑠 2 −0.1159𝑠+0.149
All models are given in Simulink in a single separate file. The input saturation boundaries and output noise are included in the file.
➢ Design PID controllers for all three transfer functions and evaluate which of the three measurements is preferred for anti-slug
control. Additionally, discuss whether the potential extra expenses and maintenance are worth the better control solution.

The control objective:

• Eliminate the oscillations fast (small settling time) and efficiently (no steady-state error), while
• Operating the system at large valve openings (𝑧𝑎𝑣𝑒𝑟𝑎𝑔𝑒 has to be as large as possible)
The three different transfer function models must be analyzed open-loop, and the controllers have to be evaluated based on the
complete closed-loop systems’ analysis. It is assumed that if one of the considered process outputs is stabilized, all other critical
pressure, temperature and flow values are stabilized as well.
• Hint: The complete controller design (e.g. consisting of a PIDF controller with root-locus gain) has to be developed in MATLAB and
Simulink, and evaluated based on both simulations in Simulink and MATLAB analysis (step responses, impulse responses, pole-
zero mapping, stability analysis etc.)
4
question 1: Information of the provided process model

• All three models are linearized around z=u=0.3. Thus, the most precise model operation is around 0.3, and that this offset has to be added
before the input saturation block in Simulink. In a similar way, the ∆𝑦 is the output of the transfer function, and thus the y(0) value has to be
added to the output signal. The complete system will be the initial input, u(0) added to the ∆𝑢, and the y(0) to the ∆𝑦 to get y. This is already
included in the Simulink model together with input saturation boundaries and output noise.
• Due to the linearization, the pressure setpoint values are not necessarily realistic, and for that reason you should focus more on how large
openings the input can operate in a stable manner. This is especially needed when the system is operated far from the linearization point
(z=0.3)
• Notice that for transfer function 2 (the topside mass flow) the reference value must be equal to 𝒚𝟐 𝟎 = 𝟎. 𝟏𝟖𝟎𝟐 because the pipeline has a
constant inflow, and thus the flow rate setpoint cannot be increased or decreased. If another setpoint is used, it will mess up the integral
term.

5
 QUESTION 2
CONTROL OF DC MOTOR WITH PENDULUM ARM

6
question 2 – Control of DC motor with pendulum arm

• a) Consider a DC motor with a pendulum arm directly attached to the motor’s shaft. Make a PIDF controller for the:
1. Angular position (𝜃𝑚 )
2. Angular velocity (𝜃ሶ𝑚 = 𝜔𝑚 )
• The controller has to be tuned using manual tuning. The main objective is to end up with a complete closed-loop
system, 𝐻 𝑠 = 𝑌(𝑠)/𝑟𝑒𝑓(𝑠), where the output’s response (to a step input) has a fast settling time, with no steady-state
error and minimum overshooting. Newton’s second law of rotation is used for the system:
(𝐽𝑚 + 𝐽𝑝 ) ⋅ 𝜃ሷm 𝑡 = σ 𝜏 = 𝜏𝑚 − 𝜏𝑓 − 𝜏𝑔
𝐾𝑡 𝐾𝑒 𝐾𝑡
• where: 𝜏𝑚 = 𝑣𝑎 𝑡 − ⋅ 𝜃ሶm (𝑡), 𝜏𝑓 = 𝑏𝜃ሶ𝑚 (𝑡), 𝜏𝑔 = 𝑚𝑔𝑟𝑠𝑖𝑛 𝜃m 𝑡
𝑅𝑎 𝑅𝑎
• The complete equation of the DC motor (with attached pendulum):
K ⋅𝐾 𝐾𝑡
𝐽𝑚 + 𝐽𝑝 ⋅ 𝜃ሷm 𝑡 + 𝑏 + t 𝑒 ⋅ 𝜃ሶm 𝑡 + 𝑚𝑔𝑟𝑠𝑖𝑛 𝜃m 𝑡 = ⋅ 𝑣𝑎 (𝑡)
𝑅𝑎 𝑅𝑎
• The pendulum moment of inertia is: 𝐽𝑝 = 𝑚𝑟 2
• The motor coefficient is equal to the electrical coefficient: K t = 𝐾𝑒
• As the pendulum is directly attached to the motor it is clear that: 𝜃m 𝑡 = 𝜃p 𝑡
Hint: Linearize the non-linearities (at hanging position) in the model equations and apply Laplace transform to the linearized
model and find the transfer functions. Use the transfer functions for obtaining the controller coefficients. Simulate the
obtained controller with the non-linear model in Simulink.

7
question 2 – Control of DC motor with pendulum arm

b) Compare the linear and non-linear models in Simulink. The comparison should be based on both open- (without
controller) and closed-loop (with controller) simulations.
Hint for open-loop linear/non-linear comparsion: Try to step the voltage input from 0 to 2, then from 0 to 10, and lastly from 0
to 15 V. Observe the differences in the output response.
c) In some cases the gravity plays an insignificant role to these types of systems. Compare the considered system with and
without the gravitational term: 𝜏𝑔 = 𝑚𝑔𝑟𝑠𝑖𝑛 𝜃m 𝑡 . Compare both as open-loop (without controller) and closed-loop (with
controller).

8
question 2 – Control of DC motor with pendulum arm

Table of parameters and coefficients:

𝑚
• g is gravitational acceration: 9.82
𝑠2
• r is the distance to the pivot of the pendulum: 0.03 m
• m is the mass of the pendulum: 0.05 kg
𝑉
• 𝐊 𝐞 is the back electromotive force coefficient: 0.05
𝑟𝑎𝑑/𝑠
• 𝑱𝒎 is the moment of inertia of the motor: 0.01 𝑘𝑔 ∙ 𝑚2
• 𝐑 𝐚 is the electrical resistance in the DC circuit: 30 ohm
• b is the friction coefficient: 0.01

9
 QUESTION 3
CONTROL OF A SEPARATION TANK

12
question 3– Control of a separation tank

We would like to control the water level in a open gas-liquid separator

tank that looks like a cylinder with cross-sectional area, 𝐴 = 2𝑚2 .
A differential equation that describes the changes of the water volume is:
𝑉ሶ = 𝐹𝑖𝑛 − 𝐹𝑜𝑢𝑡 , where V is the volume of water, 𝐹𝑖𝑛 is the water input flow
𝑚3
and 𝐹𝑜𝑢𝑡 is the outgoing water flow. 𝐹𝑖𝑛 = 0.1 is a constant.
𝑠
The output flow through a valve can be obtained from: 𝐹𝑜𝑢𝑡 = 𝐶𝑢 ℎ,
where 𝐶 = 0.25 is a constant, u is the opening percentage of the valve
and h is the water level in the tank.
1. Write the (nonlinear) dynamic model, with the water level, h, in the
tank as output and the valve opening, u, as input.
2. Draw a Simulink/block diagram for the model. Simulate the nonlinear
model in MATLAB/Simulink, giving suitable inputs, such as step.
3. Develop a PID controller to stabilize h, when the initial output is 0m.
4. Investigate the need for integral anti-windup.

13