1、 Table of Contents1. Introduction 12. Process Models 52.1 Introduction 52.2 Static Models 62.3 Dynamic Models 82.4 Step Response Methods 112.5 Methods of Moments 242.6 Frequency Responses 342.7 Parameter Estimation 432.8 Disturbance Models 462.9 Approximate Models and Unmodeled Dynamics 512.10 Concl
2、usions 572.11 References 583. PID Control 593.1 Introduction 593.2 The Feedback Principle 603.3 PID Control 643.4 Modifications of the PID Algorithm 703.5 Integrator Windup 803.6 Digital Implementation 933.7 Operational Aspects 1033.8 Commercial Controllers 1083.9 When Can PID Control Be Used? 1093.
3、10 Conclusions 1163.11 References 1174. Controller Design 1204.1 Introduction 1204.2 Specifications 1214.3 Ziegler-Nichols and Related Methods 1344.4 Loop Shaping 1514.5 Analytical Tuning Methods 156vii viii Table of Contents4.6 Optimization Methods 1644.7 Pole Placement 1734.8 Dominant Pole Design
4、1794.9 Design for Disturbance Rejection 1934.10 Conclusions 1964.11 References 1975. New Tuning Methods 2005.1 Introduction 2005.2 A Spectrum of Tools 2015.3 Step-Response Methods 2035.4 Frequency-Response Methods 2125.5 Complete Process Knowledge 2185.6 Assessment of Performance 2205.7 Examples 224
5、5.8 Conclusions 2285.9 References 2286. Automatic Tuning and Adaptation 2306.1 Introduction 2306.2 Process Knowledge 2326.3 Adaptive Techniques 2326.4 Model-Based Methods 2376.5 Rule-Based Methods 2416.6 Commercial Products 2436.7 Integrated Tuning and Diagnosis 2626.8 Conclusions 2706.9 References
6、2707. Control Paradigms 2737.1 Introduction 2737.2 Cascade Control 2747.3 Feedforward Control 2817.4 Model Following 2847.5 Nonlinear Elements 2877.6 Neural Network Control 2957.7 Fuzzy Control 2987.8 Interacting Loops 3047.9 System Structuring 3137.10 Conclusions 3217.11 References 321Bibliography
7、323Index 339IntroductionThe PID controller has several important functions: it provides feed-back; it has the ability to eliminate steady state offsets through in-tegral action; it can anticipate the future through derivative action.PID controllers are sufficient for many control problems, particula
8、rlywhen process dynamics are benign and the performance requirementsare modest. PID controllers are found in large numbers in all indus-tries. The controllers come in many different forms. There are stand-alone systems in boxes for one or a few loops, which are manufacturedby the hundred thousands y
9、early. PID control is an important ingre-dient of a distributed control system. The controllers are also em-bedded in many special-purpose control systems. In process control,more than 95% of the control loops are of PID type, most loops areactually PI control. Many useful features of PID control ha
10、ve not beenwidely disseminated because they have been considered trade secrets.Typical examples are techniques for mode switches and anti-windup.PID control is often combined with logic, sequential machines, se-lectors, and simple function blocks to build the complicated automa-tion systems used for
11、 energy production, transportation, and manu-facturing. Many sophisticated control strategies, such as model pre-dictive control, are also organized hierarchically. PID control is usedat the lowest level; the multivariable controller gives the setpoints tothe controllers at the lower level. The PID
12、controller can thus be saidto be the “bread and butter” of control engineering. It is an importantcomponent in every control engineers toolbox.PID controllers have survived many changes in technology rang-ing from pneumatics to microprocessors via electronic tubes, tran-sistors, integrated circuits.
13、 The microprocessor has had a dramaticinfluence on the PID controller. Practically all PID controllers madetoday are based on microprocessors. This has given opportunities toprovide additional features like automatic tuning, gain scheduling,and continuous adaptation. The terminology in these areas i
14、s notwell-established. For purposes of this book, auto-tuning means thatthe controller parameters are tuned automatically on demand froman operator or an external signal, and adaptation means that theparameters of a controller are continuously updated. Practically all1 2 Chapter 1 Introductionnew PI
15、D controllers that are announced today have some capabilityfor automatic tuning. Tuning and adaptation can be done in manydifferent ways. The simple controller has in fact become a test benchfor many new ideas in control.The emergence of the fieldbus is another important development.This will drasti
16、cally influence the architecture of future distributedcontrol systems. The PID controller is an important ingredient ofthe fieldbus concept. It may also be standardized as a result of thefieldbus development.A large cadre of instrument and process engineers are familiarwith PID control. There is a w
17、ell-established practice of installing,tuning, and using the controllers. In spite of this there are substantialpotentials for improving PID control. Evidence for this can be foundin the control rooms of any industry. Many controllers are put in man-ual mode, and among those controllers that are in
18、automatic mode,derivative action is frequently switched off for the simple reason thatit is difficult to tune properly. The key reasons for poor performanceis equipment problems in valves and sensors, and bad tuning prac-tice. The valve problems include wrong sizing, hysteresis, and stiction.The mea
19、surement problems include: poor or no anti-aliasing filters;excessive filtering in “smart” sensors, excessive noise and impropercalibration. Substantial improvements can be made. The incentive forimprovement is emphasized by demands for improved quality, whichis manifested by standards such as ISO 9
20、000. Knowledge and un-derstanding are the key elements for improving performance of thecontrol loop. Specific process knowledge is required as well as knowl-edge about PID control.Based on our experience, we believe that a new era of PID controlis emerging. This book will take stock of the developme
21、nt, assess itspotential, and try to speed up the development by sharing our expe-riences in this exciting and useful field of automatic control. The goalof the book is to provide the technical background for understandingPID control. Such knowledge can directly contribute to better productquality.Pr
22、ocess dynamics is a key for understanding any control problem.Chapter 2 presents different ways to model process dynamics thatare useful for PID control. Methods based on step tests are discussedtogether with techniques based on frequency response. It is attemptedto provide a good understanding of t
23、he relations between the differentapproaches. Different ways to obtain parameters in simple transferfunction models based on the tests are also given. Two dimension-free parameters are introduced: the normalized dead time and thegain ratio are useful to characterize dynamic properties of systemscomm
24、only found in process control. Methods for parameter estimationare also discussed. A brief description of disturbance modeling is alsoChapter 1 Introduction 3given.An in depth presentation of the PID controller is given in Chap-ter 3. This includes principles as well as many implementation de-tails,
25、 such as limitation of derivative gain, anti-windup, improvementof set point response, etc. The PID controller can be structured in dif-ferent ways. Commonly used forms are the series and the parallelforms. The differences between these and the controller parametersused in the different structures a
26、re treated in detail. Implementationof PID controllers using digital computers is also discussed. The un-derlying concepts of sampling, choice of sampling intervals, and anti-aliasing filters are treated thoroughly. The limitations of PID controlare also described. Typical cases where more complex c
27、ontrollers areworthwhile are systems with long dead time and oscillatory systems.Extensions of PID control to deal with such systems are discussedbriefly.Chapter 4 describes methods for the design of PID controllers.Specifications are discussed in detail. Particular attention is given tothe informat
28、ion required to use the methods. Many different meth-ods for tuning PID controllers that have been developed over the yearsare then presented. Their properties are discussed thoroughly. A rea-sonable design method should consider load disturbances, model un-certainty, measurement noise, and set-poin
29、t response. A drawbackof many of the traditional tuning rules for PID control is that suchrules do not consider all these aspects in a balanced way. New tuningtechniques that do consider all these criteria are also presented.The authors believe strongly that nothing can replace understand-ing and in
30、sight. In view of the large number of controllers used inindustry there is a need for simple tuning methods. Such rules willat least be much better than “factory tuning,” but they can always beimproved by process modeling and control design. In Chapter 5 wepresent a collection of new tuning rules th
31、at give significant improve-ment over previously used rules.In Chapter 6 we discuss some techniques for adaptation and au-tomatic tuning of PID controllers. This includes methods based onparametric models and nonparametric techniques. A number of com-mercial controllers are also described to illustr
32、ate the different tech-niques. The possibilities of incorporating diagnosis and fault detectionin the primary control loop is also discussed.In Chapter 7 it is shown how complex control problems can besolved by combining simple controllers in different ways. The controlparadigms of cascade control,
33、feedforward control, model following,ratio control, split range control, and control with selectors are dis-cussed. Use of currently popular techniques such as neural networksand fuzzy control are also covered briefly.4 Chapter 1 IntroductionReferencesA treatment of PID control with many practical h
34、ints is given in(Shinskey, 1988). There is a Japanese text entirely devoted to PIDcontrol by (Suda et al., 1992). Among the books on tuning of PIDcontrollers, we can mention (McMillan, 1983) and (Corripio, 1990),which are published by ISA.There are several studies that indicate the state of the art
35、of in-dustrial practice of control. The Japan Electric Measuring InstrumentManufacturers Association conducted a survey of the state of processcontrol systems in 1989, see (Yamamoto and Hashimoto, 1991). Ac-cording to the survey more than than 90% of the control loops wereof the PID type.The paper,
36、(Bialkowski, 1993), which describes audits of papermills in Canada, shows that a typical mill has more than 2000 controlloops and that 97% use PI control. Only 20% of the control loops werefound to work well and decrease process variability. Reasons for poorperformance were poor tuning (30%) and val
37、ve problems (30%). Theremaining 20% of the controllers functioned poorly for a variety ofreasons such as: sensor problems, bad choice of sampling rates, andanti-aliasing filters. Similar observations are given in (Ender, 1993),where it is claimed that 30% of installed process controllers operatein m
38、anual, that 20% of the loops use “factory tuning,” i.e., defaultparameters set by the controller manufacturer, and that 30% of theloops function poorly because of equipment problems in valves andsensors.Process Models2.1 IntroductionA block diagram of a simple control loop is shown in Figure 2.1. Th
39、esystem has two major components, the process and the controller, rep-resented as boxes with arrows denoting the causal relation betweeninputs and outputs. The process has one input, the manipulated vari-able, also called the control variable. It is denoted by u. The processoutput is called process
40、variable (PV) and is denoted by y. This vari-able is measured by a sensor. The desired value of the process variableis called the setpoint (SP)or the reference value. It is denoted by ysp.The control error e is the difference between the setpoint and theprocess variable, i.e., e = ysp y. The control
41、ler in Figure 2.1 hasone input, the error, and one output, the control variable. The figureshows that the process and the controller are connected in a closedfeedback loop.The purpose of the system is to keep the process variable closeto the desired value in spite of disturbances. This is achieved b
42、y thefeedback loop, which works as follows. Assume that the system is inequilibrium and that a disturbance occurs so that the process variablebecomes larger than the setpoint. The error is then negative and thecontroller output decreases which in turn causes the process outputto decrease. This type
43、of feedback is called negative feedback, becausethe manipulated variable moves in direction opposite to the processvariable.The controller has several parameters that can be adjusted. Thecontrol loop performs well if the parameters are chosen properly. Itperforms poorly otherwise, e.g., the system m
44、ay become unstable.The procedure of finding the controller parameters is called tuning.This can be done in two different ways. One approach is to choosesome controller parameters, to observe the behavior of the feedbacksystem, and to modify the parameters until the desired behavior isobtained. Anoth
45、er approach is to first develop a mathematical modelthat describes the behavior of the process. The parameters of thecontroller are then determined using some method for control design.5 6 Chapter 2 Process ModelsController Processe u y1y spFigure 2.1 Block diagram of a simple feedback system.An und
46、erstanding of techniques for determining process dynamics isa necessary background for both methods for controller tuning. Thischapter will present such techniques.Static models are discussed in the next section. Dynamic modelsare discussed in Section 2.3. Transient response methods, which areuseful for determining simple dynamic models of the process, are pre-sented in Sect
