TUSTP 2003

1
TUSTP 2003
Intelligent Control of Compact Separation System
by Vasudevan Sampath
May 20, 2003
2
Overview

• Objectives
• Literature Review
• Compact Separation System
• Review of Control System Development
• Fuzzy Logic System
• Artificial Neural Network System
• Future Plans

3
Objectives

• Conduct a detailed study on advanced control
  systems like fuzzy logic, neural network etc. and
  study their suitability for compact separation
  system.
• Develop an intelligent control strategy for
  compact separation system and conduct dynamic
  simulation and experimental investigation on the
  developed strategy.

4
Literature Review

• Control System Studies
• Wang (2000) Dynamic Simulation, Experimental
  Investigation and Control System Design of GLCC?
• Dorf Bishop (1998) Modern Control Systems
• Grimble (1994) Robust Industrial Control
• Friedland (1996) Advanced Control System Design

5
Literature Review

• Fuzzy Logic and Neural Networks
• McNeill and Thro (1994) Fuzzy Logic
• Leondes (1999) Fuzzy Theory Systems Techniques
  and Applications
• Terano, Asai and Sugeno (1994) Applied Fuzzy
  Systems
• Passino and Yurkovich (1998) Fuzzy Control
• Reznik (1997) Fuzzy Controllers

6
Compact Separation System 1
LC-Level Control PC-Pressure Control WCC-Water
cut Control FC-Feed Control PDC-Press. Diff.
Control
Oil Rich
Oil Rich
Clean
Clean
LC
LC
Pipe Type
Pipe Type
PC
PC
Separator
Separator
GLCC (Scrubber)
GLCC (Scrubber)
Manifold
Slug Damper
WCC
WCC
Water Rich
Water Rich
GLLCC (3
-
phase)
GLLCC (3
-
phase)
Clean Water
7
Compact Separation System 2
LC-Level Control PC-Pressure Control WCC-Water
cut Control FC-Feed Control PDC-Press. Diff.
Control
Gas Stream
Clean Gas
Clean
LC
LC
Pipe Type
GLCC (Scrubber)
Separator
Pump
Pump
Manifold
Slug Damper
FC
WC
PRC
PDC
LLCC
LLCC
Liquid Stream
PRC
PDC
Hydrocyclones
Hydrocyclones
Clean Water
GLCC
8
(No Transcript)
9
Control System Development Stages

• 1st Stage Frequency response design methods for
  scalar systems by Nyquist, Bode
• 2nd Stage The state-space approach to optimal
  control and filtering theory
• 3rd Stage Multivariable systems by
  frequency-domain design methods (MIMO)
• 4th Stage Robust design procedures - H? design
  philosophy
• 5th Stage Advanced techniques Fuzzy Logic,
  Neural Networks, Artificial Intelligence.

10
Adaptive Versus Robust Control

• Adaptive Control Estimates parameters and
  calculates the control accordingly. Involves
  online design computations, difficult to
  implement.
• Robust Control This allows for uncertainty in
  the design of a fixed controller, thus, producing
  a robust scheme, which is insensitive to
  parameter variations or disturbances. H? robust
  control philosophy provides optimal approach to
  improve robustness of a controlled system.

11
Limitations of Conventional Controllers

• Plant non-linearity Nonlinear models are
  computationally intensive and have complex
  stability problems.
• Plant uncertainty A plant does not have accurate
  models due to uncertainty and lack of perfect
  knowledge.
• Uncertainty in measurements Uncertain
  measurements do not necessarily have stochastic
  noise models.
• Temporal behavior Plants, Controllers,
  environments and their constraints vary with
  time. Time delays are difficult to model.

12
Fuzzy Logic Control
How are you going to park a car ?
You have to switch to reverse, then push an
accelerator for 3 minutes and 46 seconds and keep
a speed of 15mph and move 5m back after that
try..
Its eeeeassy! Just move slowly back and avoid
any obstacles.
Crisp man
Fuzzy man
13
Benefits of Fuzzy Logic Controller

• Can cover much wider range of operating
  conditions than PID and can operate with noise
  and disturbance.
• Developing a fuzzy logic controller is cheaper
  than developing a model-based controller.
• Fuzzy controllers are customizable. Since it is
  easier to understand and modify their rules.

14
Operation of Conventional Controller
Input
Output
PID Controller
PLANT
Feedback Signal
15
Operation of Fuzzy Logic Controller
Reference Input r(t)
Input u(t)
Output
PLANT
16
Fuzzy Controller Operation
17
Neural Network Process Control Loop
Input
Output
Sensing System
Plant Operating System
Neural Network Analysis System
Neural Network Decision System
18
Basic Artificial Neural Network
19
Basic Artificial Neural Network
Feed forward ANN a,b Feed back ANN - c
20
Advantages of Neural Network

• Simultaneous use of large number of relatively
  simple processors, instead of using very powerful
  central processor.
• Parallel computation enables short response
  times for tasks that involve real time
  simultaneous processing of several signals.
• Each processor is an adaptable non linear device.

21
Neuro Fuzzy Systems

• Neural Networks are good at recognizing patterns,
  not good at explaining how they reach that
  decision
• Fuzzy logic are good at explaining their decision
  but they cannot automatically acquire the rules
  they use to make those decisions
• Central hybrid system which can combine the
  benefits of both are used for intelligent systems
• Complex domain like process control applications
  require such hybrid systems to perform the
  required tasks intelligently
• In theory neural network and fuzzy systems are
  equivalent in that they are convertible, yet in
  practice each has its own advantages and
  disadvantages

22
Applications

• Fuzzy Logic and Neural Network applications to
  compact separation system
• Dedicated control system for each component,
  like GLCC or LLCC
• Sensor fusion improvement in reliability and
  robustness of sensors
• Supervisory control intelligent control system
  with diagnostics capabilities.

23
Future Plans

1. Develop dedicated control systems for each
   component using neural network or adaptive
   control system.
2. Develop sensor fusion modules using neural
   networks to improve the quality of measured
   signal.
3. Develop intelligent supervisory control system
   for overall control, monitoring and diagnostics
   of the process.