CONTROL SYSTEM DESIGN - University of Alabamafeihu.eng.ua.edu/NSF_CPS/year1/w7_2.pdf · Chapter 1...
Transcript of CONTROL SYSTEM DESIGN - University of Alabamafeihu.eng.ua.edu/NSF_CPS/year1/w7_2.pdf · Chapter 1...
CONTROL SYSTEMCONTROL SYSTEMDESIGNDESIGN
Graham C. GoodwinStefan F. Graebe
Mario E. Salgado
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Chapter 1
The Excitement of ControlThe Excitement of ControlEngineeringEngineering
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Motivation for Control Engineering
Feedback control has a long history which beganwith the early desire of humans to harness thematerials and forces of nature to their advantage.Early examples of control devices include clockregulating systems and mechanisms for keepingwind-mills pointed into the wind.Modern industrial plants have sophisticated controlsystems which are crucial to their successfuloperation.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
A modern industrial plant: A section ofthe OMV Oil Refinery in Austria
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Control Engineering has had a major impact onsociety. For example, Watt’s Fly Ball Governor hada major impact on the industrial revolution. Indeed,most modern systems (aircraft, high speed trains, CDplayers, … ) could not operate without the aid ofsophisticated control systems.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Figure 1.1: Watt’s fly ball governor
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
This photograph shows a flyballgovernor used on a steam enginein a cotton factory near Manchesterin the United Kingdom. Of course,Manchester was at the centre of theindustrial revolution. Actually, thiscotton factory is still running today.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
This flyball governor is in the same cotton factory in Manchester.However, this particular governorwas used to regulate the speed ofa water wheel driven by the flow ofthe river. The governor is quite large as can be gauged by the outlineof the door frame behind the governor.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Improved control is a key enabling technologyunderpinning:❖ enhanced product quality❖ waste minimization❖ environmental protection❖ greater throughput for a given installed capacity❖ greater yield❖ deferring costly plant upgrades, and❖ higher safety margins
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Figure 1.2: Process schematic of a Kellogg ammonia plant
All of the above issues are relevant to the control of an integrated plant such as that shown below.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Types of Control System Design
Control system design also takes several differentforms and each requires a slightly differentapproach.The control engineer is further affected by wherethe control system is in its lifecycle, e.g.:❖ Initial "grass roots" design❖ Commissioning and Tuning❖ Refinement and Upgrades❖ Forensic studies
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
System Integration
Success in control engineering depends on taking aholistic viewpoint. Some of the issues are:❖ plant, i.e. the process to be controlled❖ objectives❖ sensors❖ actuators❖ communications❖ computing❖ architectures and interfacing❖ algorithms❖ accounting for disturbances and uncertainty
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Plant
The physical layout of a plant is an intrinsic part ofcontrol problems. Thus a control engineer needs tobe familiar with the "physics" of the process understudy. This includes a rudimentary knowledge of thebasic energy balance, mass balance and materialflows in the system.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Objectives
Before designing sensors, actuators or controlarchitectures, it is important to know the goal, that is,to formulate the control objectives. This includes
❖ what does one want to achieve (energy reduction, yield increase,...)
❖ what variables need to be controlled to achievethese objectives
❖ what level of performance is necessary (accuracy, speed,...)
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Sensors
Sensors are the eyes of control enabling one to seewhat is going on. Indeed, one statement that issometimes made about control is:
If you can measure it, you can control it.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Actuators
Once sensors are in place to report on the state of aprocess, then the next issue is the ability to affect, oractuate, the system in order to move the processfrom the current state to a desired state
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
A typical industrial control problem will usuallyinvolve many different actuators - see below:
Figure 1.3: Typical flatness control set-up for rolling mill
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
A modern rolling mill
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Communications
Interconnecting sensors to actuators, involves the useof communication systems. A typical plant can havemany thousands of separate signals to be sent overlong distances. Thus the design of communicationsystems and their associated protocols is anincreasingly important aspect of modern controlengineering.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Computing
In modern control systems, the connection betweensensors and actuators is invariably made via acomputer of some sort. Thus, computer issues arenecessarily part of the overall design. Currentcontrol systems use a variety of computationaldevices including DCS's (Distributed ControlSystems), PLC's (Programmable Logic Controllers),PC's (Personal Computers), etc.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
A modern computer based rapid prototyping system
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Architectures and interfacing
The issue of what to connect to what is a non-trivialone in control system design. One may feel that the bestsolution would always be to bring all signals to acentral point so that each control action would be basedon complete information (leading to so called,centralized control). However, this is rarely (if ever) thebest solution in practice. Indeed, there are very goodreasons why one may not wish to bring all signals to acommon point. Obvious objections to this includecomplexity, cost, time constraints in computation,maintainability, reliability, etc.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Table 1.1: Typical control heirarchy
Level Description Goal Timeframe
Typicaldesign tool
4 Plant wide opti-mization
Meeting customer orders andscheduling supply of materials
Everyday(say)
Static opti-mization
3 Steady state op-timization at unitoperational level
Efficient operation of a singleunit (e.g. distillation column)
Everyhour(say)
Static opti-mization
2 Dynamic control atunit operation level
Achieving set-points specifiedat level 3 and achieving rapidrecovery from disturbances
Everyminute(say)
Multivariablecontrol, e.g.Model Predic-tive Control
1 Dynamic controlat single actuatorlevel
Achieving liquid flow rates etcas specified at level 2 by ma-nipulation of available actua-tors (e.g. valves)
Everysecond(say)
Single variablecontrol, e.g.PID
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Algorithms
Finally, we come to the real heart of control engineeringi.e. the algorithms that connect the sensors to the actuators.It is all to easy to underestimate this final aspect of theproblem.As a simple example from our everyday experience,consider the problem of playing tennis at top internationallevel. One can readily accept that one needs good eye sight(sensors) and strong muscles (actuators) to play tennis atthis level, but these attributes are not sufficient. Indeedeye-hand coordination (i.e. control) is also crucial tosuccess.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
In summary:
Sensors provide the eyes and actuators the musclebut control science provides the finesse.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
� Better ControlProvides more finesse by combining sensors and actuators in more intelligent ways
� Better ActuatorsProvide more Muscle
� Better SensorsProvide better Vision
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Disturbances and Uncertainty
One of the things that makes control scienceinteresting is that all real life systems are acted on bynoise and external disturbances. These factors canhave a significant impact on the performance of thesystem. As a simple example, aircraft are subject todisturbances in the form of wind-gusts, and cruisecontrollers in cars have to cope with different roadgradients and different car loadings.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Homogeneity
A final point is that all interconnected systems,including control systems, are only as good as theirweakest element. The implications of this in controlsystem design are that one should aim to have allcomponents (plant, sensors, actuators, communications,computing, interfaces, algorithms, etc) of roughlycomparable accuracy and performance.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
In order to make progress in control engineering (asin any field) it is important to be able to justify theassociated expenditure. This usually takes the formof a cost benefit analysis.
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Cost benefit analysis
Typical steps include:❖ Assessment of a range of control opportunities;❖ Developing a short list for closer examination;❖ Deciding on a project with high economic or
environmental impact;❖ Consulting appropriate personnel (management, operators,
production staff, maintenance staff etc.);❖ Identifying the key action points;❖ Collecting base case data for later comparison;❖ Deciding on revised performance specifications;❖ Updating actuators, sensors etc.;
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
❖ Development of algorithms;❖ Testing the algorithms via simulation;❖ Testing the algorithms on the plant using a rapid
prototyping system;❖ Collecting preliminary performance data for comparison
with the base case;❖ Final implementation;❖ Collection of final performance data;
❖ Final reporting on project.
Cost benefit analysis (Contd.)
Goodwin, Graebe, Salgado ©, Prentice Hall 2000Chapter 1
Signals and systems terminology
Tangible examples Examples of mathematicalapproximation
Examples of properties
Signals set point, controlinput, disturbances,measurements, ...
continuous function, sample-sequence, random process,...
analytic, stochastic, sinu-soidal, standard deviations
Systems process, controller,sensors, actuators, ...
differential equations, differenceequations, transfer functions, statespace models, ...
continuous time, sampled,linear, nonlinear, ...