The century's greatest contributions to control practice
ISA Transactions 39 (2000) 3±13 www.elsevier.com/locate/isatrans Editorial viewpoint The centurys greatest contributions to control practice R. Russell Rhinehart* School of Chemical Engineering, Oklahoma State University, Stillwater, OK 74078-5021, USA Flowing with the tide of re¯ection and forecast- 1. Survey participantsing as the century changes, I took the liberty toconduct a survey of the centurys most important The participants included about 100 profes-contributions to the practice of ``control. I quite sionals, chosen to represent both industry andenjoyed considering the importance of the past academe, to represent a wide range of engineeringevents, and forecasting which contributions will be disciplines (Aeronautical, Chemical, Computer,important in this new century. So did many of the Electrical, Instrument, and Mechanical), and tosurvey respondents. I hope that you will enjoy this represent a wide range of perspectives (vendors,re¯ection, and also ®nd it useful. users, researchers, developers, and managers). This was a non-scienti®c survey, with a limited This balance of people mainly came from thosenumber of participants. However, the participants who are active in ISA Transactions and theare all established experts, who represent a wide American Automatic Control Council. From thevariety of disciplines and professions, and who academics, I selected those with a strong applica-appear to have submitted well-considered opi- tions orientation. From industry folks, I selectednions. Accordingly, I believe the results are fairly those working at the leading edge. I classify thelegitimate, and the diversity and number of parti- participants as either practical academics orcipants makes the results free of marketing bias. advanced practitioners. For the most part, the For those who are directing the development of academics and practitioners had similar views oftextbooks, handbooks, degree programs, continu- the relative importance of the contributions.ing education programs, top elements in the sur- I also selected senior colleagues, so that theirvey will be useful for specifying content. Those retrospective on the century will be more com-same elements will be a useful guide for those plete. However, even the most senior gurus ofdirecting their own professional development. For today only experienced their profession in the lastthose who are planning and managing the techni- third of the century. This creates a perspectivecal future of an operating company, a vendor, or a bias. For example, when using symbolic pro-service provider, this survey contains a useful cessors and spreadsheet macros, it is hard tovision of the most promising and practicable new appreciate the signi®cance of the slide-rule. It dis-technology. appeared about 25 years ago; but until then, was an indispensable tool. Many of us never experi- enced its importance; and even for those of us who * Tel.: +1-405-744-5280; fax: +1-405-744-6338. did, the computer has trivialized our remembrance E-mail address: email@example.com of the slide-rules importance. Respondents were0019-0578/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.PII: S0019-0578(00)00010-0
4 R.R. Rhinehart / ISA Transactions 39 (2000) 3±13instructed to be aware of this; but even so, this process, or they manipulate the process.survey must have a bias for the last quarter of the They include hardware as well as software.century. 4. Control, Decision, and Communication Instrumentation Ð These contributions, either hardware or software, are the ones that2. Survey categories implement the control strategy and automate the decision. The survey started with a small, balanced group 5. Organization Ð These items included orga-of people to help create both the scope and the nizations of people (such as ISA and NASA)de®nition of ``the practice of control. This would that contributed to control practice by stan-guide our selection of the most signi®cant con- dardizing, demonstrating, and disseminatingtributions, which come from a variety of distinctly techniques. The list also includes books, andierent directions. Theory contributes to practice organization of technology that was sig-when it directly leads to a tool that comes into ni®cant to acceptance and use.widespread use. Top examples included frequencyanalysis that led to rules for controller designing Considering this scope of contributions, theseand tuning, and state space analysis that led to statements guided selection for the ``most sig-modern control techniques and the Kalman ®lter. ni®cant contributions to the practice of control:Instrument devices were included, with some being Include all aspects of ``control Ð from sensingthe pneumatic-mechanical PID wonder, the gas to system optimization, from technology tochromatograph, the Norden bombsight, the people. To focus on the practice, consider wherepneumatic valve actuator, and the Coriolis ¯ow- non-control specialists use ``control Ð where itmeter. Of course, the computer and communica- is a useful tool to make something work. Totion systems were included as tools that ®nally focus on the most important contributions, con-allowed data access, the automation of process sider both the tangible bene®t to human welfareanalysis and higher level logic, nonlinear control, that was achieved, the degree of acceptance byand optimization. Control strategies such as mul- those who use control as a tool, the paradigmtivariable model predictive control and fuzzy logic shift that resulted from the contribution, or thecontrol were included. Finally, organizational magnitude of the implementation need that wasevents such as textbooks and professional societies satis®ed.that revealed and standardized the practice of The ``demonstrated utility aspect of the criteriacontrol were added. Contributions came from a creates a bias against contributions of the past 5 towide range of in¯uences. We developed a list of 99 15 years that are just now developing into sig-items. It was a great century! ni®cance. So, we created a sixth category: Considerations of the diversity of items on thelist led to grouping of the contributions into 5 6. Most Promising Newcomers Ð These itemscategories: included ones that were introduced later in the twentieth Century, which have consider- 1. Concepts and Theory Ð These contributions able promise and signi®cant isolated demon- were signi®cant because they formed the strations, but which have not had the time to basis for techniques and products that made come to full fruition within the practice. signi®cant impact on the practice of control. 2. Techniques Ð These items are the procedures and algorithms for applying, designing, or 3. Survey stages analyzing elements of the control system. 3. Process Instrumentation Ð These items are The survey had three phases. First, about 10 of the ones that ``touch the process. They us de®ned the practice of control. Then I sent the provide or convey information about the de®nition and some starter examples to about 100
R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 5colleagues and asked them to suggest more items. perhaps should have been placed within the New-About 35 people responded. I used the initial comers category. Interestingly, it was one of a fewsmall group to help re®ne and organize the items. items where there was a moderate dierence in theTable 1 shows the entire list. Finally, I sent that importance attributed by academics and practi-list to each of the 100 and asked them to choose tioners. I have spent half of my career in industrytheir top 3±5 in each category. There were about and half in academe, and am interested in view16 items in each category, some of which might point dierences of the two careers. Academicsdominate the others; so, allowing 3±5 choices in rated Internal Model Principle high. I will let youany category would permit strong acknowl- decide whether this is a useful technology that hasedgment of the second-place items. Forty-two not been fully discovered by industry or an analy-folks returned feedback in time for this editorial to tically rich concept with little utility. I believe thego to press. Table 1 lists the raw data. Fig. 1 shows issue is worth investigating.that the mix was 44% academe and 56% industry. Fig. 3 displays the top items in the TechniquesI appreciate all of those who participated. category. In my view each of these items de®nes skills that a control engineer should have, and items on this list will continue to be important in4. Survey results the twenty-®rst century. This list should guide training and educational programs. If votes/per- Figs. 2±7 display items in each category that son is an indication, this was the most importantcollectively received about 75% of the votes. Items category to the practice of control. Instructionsthat received only a few votes were deemed to be were to ``choose 3±5 items from each list; how-signi®cant by a few people. They may even form ever, individuals averaged choosing 5.4 items perthe basis for a strong segment of the industry, but person in this category! Thats right, many peoplethey did not share the collective, widespread ar- chose 6 items. Some, 7 items. Only a few sub-mation. Shown in Fig. 2, the top items in Category mitted just three.1, Concepts and Theory were the classic feedback The item on Fig. 3 called ``Classical Advancedprinciple (bias some aspect of the control action Control includes the 30s and 40s SISO techni-with the error and a cumulative sum) and fre- ques of ratio, cascade, feedforward, override,quency analysis. These originally supported the decouplers, and reset feedback. Industry folksSISO PI control of the 20H s that continues as the rated them very important (second most, from thebackbone of control today. The importance of the list of 20 items); academics did not (12th on theirfeedback principle is evident in its number of list). For most of this study the industry and aca-votes, 40, re¯ecting that 95% of the respondents demic views were consistent. However, this waschose it as one of their top 3±5 from the list of 11 one of the two cases of statistically signi®cantitems. The other top items in the Concepts and (a=0.05) dierences between academic andTheory category are those that structure MIMO industry opinions on the 99 items in this survey. Itcontrol of the computer age: State-Space Analysis, appears that these SISO techniques will continueDigital Control Theory, and Optimal Control to be fundamental to process automation. TheyTheory. If votes per person per item is an indica- comprise the majority of ``advanced controltion of importance, this was the most important loops today. Industry is ``slow to upgrade whencategory at 0.39 votes/person/item. technology at hand in older equipment is fully Just short of making the top of the list was sucient, cost eective, and understood by``Internal Model Principle. In my view, Internal operators. Since these SISO techniques are stan-Model Control is a useful SISO tool. From its dard oerings on contemporary DCS and PLCuni®ed view, the concept leads to PID, Smith pre- equipment, they are guaranteed to be a workhorsedictor, and other control strategies. It has also of process control well into this century. Perhapslead to the ``lambda tuning methods. It is an 80H s the vote discrepancy is simply a re¯ection of respon-development within the academic community, and dent demographics. The industry respondents were
6 R.R. Rhinehart / ISA Transactions 39 (2000) 3±13Table 1Survey results on items suggested as having the greatest contributions to the practice of control in the twentieth century Number of Number of Total academe votes industry votes votesCategory 1 itemsVariational calculus 1 1 2International practical temperature scale 0 3 3Reliability methods for safety 0 4 4Strategy design by ®rst-principles models 2 8 10Internal model principle 7 4 11Least squares system identi®cation 4 7 11Optimal control 8 9 17Digital control theory 8 11 19State space and modern control theory 15 15 30Frequency response analysis 15 19 34Feedback principle 17 23 40Category 2 itemsSelf-tuning regulators 0 1 1Discrete event approaches 0 1 1Envelope control 0 2 2Cohen & Coon tuning 0 3 3Predictive maintenance/reliability-centered maintenance 0 3 3Heuristic tuning rules 1 5 6Bristols relative gain array 3 3 6Deadtime compensators 4 3 7Valve sizing 0 8 8Dynamic programming 3 5 8Statistical process control 2 8 10Linear-quadratic optimal control 6 6 12Classical advanced control 2 14 16Nyquist stability criterion 10 7 17Root locus design 9 9 18Simulation, CAD 6 13 19Multivariable MPC 7 13 20Bode design 11 9 20Kalman/kalman-Bucy ®lter 11 11 22Zeigler-nichols tuning 9 17 26Category 3 itemsBailey boiler meter 0 0 0Ultrasonic level sensors 0 0 0pH sensor/transmitter 0 1 1Nuclear level and density sensing 0 1 1Miniaturization of pneumatic instruments 2 0 2Weigh cells 0 2 2Miniature electronic sensing instruments 1 1 2Toxic and hazardous gas detectors 0 2 2Flame scanners 0 3 3Bimetallic thermostat 1 2 3Bourdon tube 3 1 4Float control 4 3 7Equal % valves 3 4 7Coriolis ¯owmeter 2 6 82-Wire control loop 3 6 9I/P transducer 2 7 9Composition analysis 1 9 10Wireless technology 4 6 10 (Table continued on next page)
R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 7Table 1 (continued) Number of Number of Total academe votes industry votes votesRobots 6 4 10Thermostat 7 3 10Gyroscope 7 4 11Fiber optics 6 7 13Valve positioners 4 10 14dp cell 2 13 15Pneumatic valve 4 17 21Category 4 itemsMiniature electronic controllers 0 1 1SCADA 0 3 3Norden bombsight 1 2 3Centralized control 2 1 3Logic solvers 1 2 3Passive zenier barrier devices for intrinsic safety 0 3 3Analogue computers 2 2 4Sequential automation 0 4 4Watt regulator 2 2 4Voting architecture 0 4 4DDC 4 3 7CRT displays for HMI 3 7 10Distributed control 3 9 12Op-Amp (operational ampli®er) 9 11 20Programming languages 5 16 21Transistor 6 19 25PID controller 15 20 35Microprocessor 17 24 41Category 5 itemsWilkinsons book 1 3 4Danzigs book 2 5 7Eckmans book 5 5 10Blacks book 3 10 13Weiners book 6 8 14Shinskys book 5 11 16P& id Symbol standards (ISA and SAMA) 5 12 17IFAC 12 5 17IEEE 10 10 20ISA 9 13 22NASA 8 14 22Category 6 itemsAuto tuning by relay feedback 2 2 4Sliding mode (variable structure control) 3 4 7Enterprise-level management systems 2 6 8Optical computing 3 5 8Connectivity tools, networks 1 8 9Nonlinear model predictive control 6 7 13Robust inferentials 5 9 14Soft sensors 3 12 15Object oriented programming 5 10 15Open systems 6 10 16Fuzzy logic 11 7 18Adaptive, self-tuning controllers 8 11 19Intelligent systems 10 10 20Neural networks 10 11 21
8 R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 Traditional control theory places the dynamics and faults of sensors and ®nal control elements within the mysterious ``plant. But, Process Instruments containing the most number of items in this survey suggests that they are very impor- tant. Perhaps we should explicitly account for the in¯uence of these items in our analysis of control systems. Fig. 1. Distribution of respondents. Fig. 5 presents the top items in the instruments used for control, decision, and communication. The ``microprocessor/computer was voted as thepredominantly from the process industry where most important contribution to control from thisthese techniques are particularly important, while category, and considering that it received 41 votesacademic respondents represented a wider variety out of 42 voters, it is the most highly rated itemof disciplines. These data cannot be de®nitive, but from this entire study. It is the tool that allows allthey suggest that the academic community should of the advanced practices in control, communica-consider the value of ``Classical Advanced Con- tions, optimization, smart instruments, etc. Attrol within the engineering curriculum. least a few of the respondents included with it the Fig. 4 shows the top items in the Process mass storage devices that are integral to the com-Instrumentation category. I believe that the vari- puter. This is one of the few items in this studyety of items and the mixture of applications that was not originally created for control. Oddly,represented on Fig. 4 arm the breadth of dis- it has become the most important. Considering theciplines participating in this survey. This category computer applications throughout the other seg-had the most number of items, 25, perhaps indi- ments of our life, the ``microprocessor/computercating the importance of the contribution of sen- could become known as the greatest contributionsors and ®nal control elements to control. to all human endeavors of the century. Fig. 2. Top concepts and theory (lower bar academe, upper bar industry).
R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 9 While ``Programming Languages and the was ``Programming Languages Ð high level over``Transistor were not as unanimous a choice as machine instruction. I had not considered pro-the microprocessor, they are of signi®cant impor- gramming languages as a major contribution totance. And, like the computer, neither was created the practice of control in my voting, and chosespeci®cally for control. The full name for voting other items. But, with the collective opinion of so Fig. 3. Top techniques (lower bar academe, upper bar industry). Fig. 4. Top process instruments (lower bar academe, upper bar industry).
10 R.R. Rhinehart / ISA Transactions 39 (2000) 3±13many gurus, in retrospect I have to agree that high challenge our views against the collective opinionslevel language has permitted the use of simulators, presented by this body of experts.on-line analysis, custom displays, fuzzy logic, and Note that the PID controller was a strong sec-much more. Certainly, it should be rated as a sig- ond on Fig. 5. When I ®rst saw one, the mystifyingni®cant contribution. Perhaps all of us could pneumatic-mechanical workings, the elegant brass Fig. 5. Top control decision and communication instruments (lower bar academe, upper bar industry). Fig. 6. Top organizations Ð of people and information (lower bar academe, upper bar industry).
R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 11bellows and levers, the clock-works delicacy of its physical measurement from the operator interfaceinternals, simplicity of drivers, and clever device. Since the operator interface no longer hadmechanisms for display and tuning immediately to support the rugged measuring device (bourdonenchanted me. Regardless of the romance of such tube, thermal system spiral, etc.), the operatordevices, and regardless of their (almost) replace- interface could be reduced in size to typicallyment by analogue or electronic equivalents, the 6HH HÂ3HH W. This permitted a much greater densitypneumatic PID instrument moved control from of information display, up to 48 variables per footgovernors (with manual bias adjustment to of panelboard. This greatly increased the numberremove oset) to truly automatic. Their contribu- of variables, therefore, the scope of the process,tion to the practice was momentous. that one operator could monitor. Some installa- Harold Wade, of Wade Associates, adds that tions incorporated a graphic diagram of the pro-the miniaturization and standardization of the cess into the control panel, with miniaturepneumatic devices was also a signi®cant event: indicators/recorders/controllers placed symboli-``Miniaturization freed the industry from reliance cally in the proper relation to the point of mea-on directly connected devices for measurement, surement of the process variable. Control roomsindication, recording, and control. Prior to this became more centralized, resulting in better coor-most control rooms consisted of a control panel with dination of information between process units.one or two rows of large case recorder-controllers, Direct Digital Control (DDC) fell just short ofnear to and directly connected to the process. A making an appearance in Fig. 5. It was in the toptypical instrument size was 16HH HÂ12HH W; each cumulative 80%, but I only showed the top 75%instrument provided one to three measurement in the ®gures. As I recall, DDC was the ®rst com-pens. This allowed an operator to monitor 4±6 mercial entry of the computer into process con-variables per linear foot of panelboard length. trol. It was short lived, as the evolved PLC andThis greatly restricted the scope of the process for DCS products overcame reliability and conceptwhich one operator could maintain surveillance. problems. It was an experiment that did not quitePneumatic signal transmission (®nally standar- work, and I suspect that the short life and rela-dized at 3±15 psig) permitted separation of the tively primitive technology of DDC resulted in its Fig. 7. Most promising newcomers (lower bar academe, upper bar industry).
12 R.R. Rhinehart / ISA Transactions 39 (2000) 3±13relatively low ranking on this list. Here, ``demon- Standards, which uni®ed communication amongstrated utility was a measure of rank, and it practitioners, was highly rated as a signi®cant con-seemed to overshadow ``paradigm shift. Cer- tribution. Statistically, the most signi®cant dier-tainly, however, the vision and bold moves of ence of opinion between academic and industrythose who accepted the DDC deserve respect. folks in this survey was on the importance of the Along the lines of that analysis, all of the items International Federation of Automatic Control.listed in Table 1, not just those making the ®gures, Again demographics could have been the reason.have legitimate claim to being substantial con- Academics were distributed around the world.tributions to control in the twenthith Century. Industry folks were almost exclusively from North Fig. 6 presents the top items on the organization Ð America. However, this may signal a message,of people and information list. The US National ``Keep an eye on IFAC, to industry.Aeronautics and Space Administration (NASA) ``Control has its landmark books; those valuedwas the top vote getter. This also surprised me. In as such a useful guide to technology that they ele-my process engineering view, I do not see con- vated the practice. The respondents placed F. G.tributions from the space programs. However, as Shinskeys book ``Process Control Systems, andAl Comello commented on his votes, NASA ``F F F Nobert Wieners book, ``Cybernetics, or Controlgave a stage for applying control concepts and a and Communication in Animal and Machines onplatform for the birth of sensors/devices that did the top of the list. If you have a copy of one ofnot exist. Industry and academia could not have these, or any of the books listed in Table 1, betterhoped to deliver this under an economic model. keep it. Museums may want them!NASA provided a marketplace and a relatively Results from the ®nal category, ``Most Promis-universal access to the technology. Al is a mem- ing Newcomers, are presented in Fig. 7. Each ofber of the Editorial Advisory board for ISA these items has a record of credible practice appli-Transactions. cations. For each there are several companies In retrospect it is odd that business organiza- selling products. These are not ``futuristic imagi-tions were not included on the list. Private enter- nations, but recent technology, fully commercialprise invested substantial development eort to today, deemed to have great potential, but notcreate products that implemented the techniques. enough time to establish itself as a ``greatest con-They brought to us the instruments and the com- tribution. These items may become known as themercial software products for data processing, twentieth Centurys great legacy for the twenty-control, and optimization. I suspect that none of ®rst. I believe that we should look at these items asus who created the survey were thinking along the we plan for the future.lines that allowed us to ``see the contributions of ``Neural Networks (NN), voted top, is a ¯ex-the commercial organizations during the brief ible method for empirical modeling. In the 80H s,period of creating the list. As I consider the mag- hype-driven, often inappropriate applications ofnitude and diversity of products, I have to apol- primitive NNs gave NNs a bad initial reputation.ogize for not including business names on the However, note that both the academic and indus-survey. However, I am somewhat relieved at the trial participants of this survey placed NNs as aoversight, for there are too many to mention most promising newcomer to the practice of con-within the page limit of this article. trol. Since NNs are often the tool that is the basis Note that the bar heights in Fig. 6 are fairly of other top items in this category (Nonlinearuniform. It seems that many organizations are Model Predictive Control, Soft Sensors or Virtualperceived as having made fairly equivalent con- On-Line Analyzers, Robust Inferentials, andtributions. The professional societies of ISA, Intelligent Systems), perhaps the importance ofIEEE, and IFAC certainly have contributed by NNs should be elevated.dissemination of useful information through con- ``Intelligent Systems are software programs thatferences, exhibitions, continuing education cour- ``learn and ``modify their action. So do ``Adap-ses, and standards. Notable, the item ``Symbol tive and Self-Tuning Controllers. In a too-simple,
R.R. Rhinehart / ISA Transactions 39 (2000) 3±13 13they-will-jump-on-me-description, Ill oer the that it oers insight for further contemplation. Ifollowing description of the dierence. Intelligent hope that it has had the same impact for you.Systems use ``unstructured non-equation-based The list in Table 1 is not complete. For instancetechniques such as NNs and genetic algorithms, the recent and important advances in batch man-and are often called agents. Presently, their main agement automation, continuous improvement,demonstrations are in games and computer- system reliability, the Internet, and data managementagainst-computer contests, but success is evident. were not included. Survey respondents pointedThe extension to process management is obvious. this out to me in their votes. I am looking forward toAdaptive and Self-Tuning Controllers are equa- see how it all sorts out in the twenty-®rst century.tion-based. The adaptation is very structured.They, too, have substantial proven applications. ``Fuzzy Logic was high on the list in spite of arelatively low number of industrial votes. In myview Fuzzy Logic Control (FLC) oers a relativelyeasy way to get a computer to take the sameheuristic action that an experienced engineer oroperator would take. Often the human intuitiveintervention (®ne-tuning, control, balancing,model adjustment, or optimization) of a system isbetter (cheaper, more robust, eective) than anyattempt by a model-based system. FLC has amultitude of successes within both the processindustry and in the OEM. Why then is there adiscrepancy in the academic and industrial per-ception of its importance? Perhaps a reason is jar- Dr. R. Russell Rhinehart is the Head of thegon. Computer and human cognitive scientists School of Chemical Engineering at Oklahomadeveloped fuzzy logic, and they must have enjoyed State University, and has considerable experiencecoming up with names like ``universe of discourse in both industry (13 years) and academe (14to describe the simple concept of ``measurement years). Russ primary research interests are in therange. I believe that the language of FLC is a practical application of advanced technology forbarrier to its application. Further, since FLC pro- automated process management (control, optimi-grams are relatively simple to write from scratch, zation, monitoring). Necessarily, his program hasFLC may not have the promotional push by ven- a strong experimental component.dors to generate industrial acceptance. Alter- Russ is an active member of AIChE and ISAnately, industry may be aware of practicable and holds several professional society positionsproblems with FLC that academe is not. We including Editor-in-Chief of ISA Transactions,should work to learn of each others view and and General Chair for the 2002 American Controleither use it, or place it lower on the list. Conference. His 1968 B.S. in Chemical Engineer- ing and M.S. in Nuclear Engineering are both from the University of Maryland. His 1985 Ph.D.5. Closing in Chemical Engineering is from North Carolina State University. Here was an unscienti®c study from an uncon- Dr. Rhineharts locators are: oce mail 423trolled demographic mix presented in an editorial, Engineering North, Oklahoma State University,not a refereed paper. It is simply one opinion of Stillwater, OK 74078-5021; Tel.: +1-405-744-5280;the most important contributions to control of the fax +1-405-744-6338;twentieth century. However, I enjoyed the exercise E-mail address: firstname.lastname@example.org;as a review of where we have been, and suspect web http://www.cheng.okstate.edu.