- Email: [email protected]
Digital computer applications to process control
Autoraatica. Vol. 14, pp. 289 291 © Pergamon Press Ltd. 1978. Printed in Great Britain
IFAC Report Digital Computer Applications to Process Control Summary of round-table discussions, 5th IFAC/IFIP International Conference; The Hague, The Netherlands, June 1977. O.
RADEMAKER*
Let us linger for a moment in 1964, when the first Computer Control conference took place in Stockholm. We were then talking mainly about registers, bits, instruction sets, poor computer reliability, and how to install computers. We worried about shielding and earth connections. We discussed optimal programming of a computer whose working memory was on a drum, as well as the feasibility of adc times smaller than 200 ms, and the reliability of relay multiplexers. Further, we speculated on the break-even number of ddc loops, wondering whether computer control would ever become really economically justifiable in view of the fantastically high price and unwieldy size of computers, and so on and so forth. Now look at 1977: True, of the 34 papers on methods only about three were not from Universities or Research Institutes, but there were many more papers looking back upon real applications:
Summary--A summary is given of the issues and conclusions of the round-table discussions on 'Economic evaluation of computer control', 'Ergonomic aspects', "Centralised versus distributed computer systems', 'Microcomputers', 'Standardisation efforts in process control software', 'Education in digital computer-control', and "Application of "Modern Control Theory" to Computer Control in the Process Industry'. Mr. Chairman, Ladies and Gentlemen, DURING this IFAC/IFIP conference, which is now almost behind us, one speaker remarked that automatic feedback started in the 1930's with the advent of electronics, but actually, it was already used a few thousand years ago by Ktesibios of Accra, who was the teacher of Heron of Alexandria, and probably even earlier by the Sumerians, so I hope the audience will not mind if I go back only about two hundred years for a minute. Back to the famous composer Haydn, who lived at a time when windmills and watermiils already had automatic control devices, and who wrote the Farewell Symphony (Abschiedssymphonie), also called The Candle Overture. In January 1772, his noble master Esterhazy moved his court from delightful Vienna to his lonesome winter residence and told his orchestra, conducted by Haydn, to leave their wives at home, because they were always quarrelling anyway. After a while, the members of the orchestra did not like that any longer and started to complain. Haydn then wrote this particular symphony, in which--one after the other--the musicians finished their part of the music, blew out their candles, picked up their instruments, and left, until the podium was empty. His Royal Highness understood this ungentle hint and said: 'Tomorrow we shall leave'. The memory of this story came back to me again when the organisers of the conference asked me to summarise the conclusions of the round-table discussions, adding that this kind of final meeting was bound to be somewhat disturbed by participants leaving to catch their plane, train, or wife, in time. So I decided to bring a candle myself, which made me remember another story: In a small area in our country, parents of an attractive girl had a very peculiar custom, centuries ago, which is now completely forgotten in Holland, but has been kept alive by descendants of Dutchmen who left the country centuries ago; kept alive, almost at the end of the world: If a beautiful girl goes steady and her boy friend visits her home, her father and mother go to sleep at some decent time, leaving a burning candle behind. Before the candle goes out, the boy is supposed to be out of the house (which goes to tell you how old this story is!). Now, the point is, what kind of candle do the parents leave behind ? If it is thin and short, they do not think much of the boy. If it is thick and tall, they like him. So I considered the request of the organisers, and since they did not give me a candle I have brought my own (a thin and short one).
41 application papers, 11 case histories, 10 survey papers or tutorials, and many of the old issues have virtually disappeared. Some have remained, as became clear in the eight round-table discussions the organisers asked me to talk about, using the reports by the chairmen as the main sources of information. Well, first of all, they were not round at all. Those I have witnessed were characterised by: - - a large panelled room full of listeners, - - a panel behind the table, and - - a variety of threads of discussion that could not be woven into a fabric within an hour. But I do think the round-table discussions left many participants with many questions to be discussed among themselves, and this conference, this building and above all you, the participants, offered splendid opportunities for doing that, so I have the impression that, bye and large, the round-table discussions have been quite successful in the overall framework of this Conference. The chairmen of the various round-table discussions have been invited to summarise their findings on a single sheet of paper. Most of them have done so, in a variety of ways. I considered the possibilities of functioning as a cross-compiler but decided against that idea, not because those chairmen might become cross with me (which is likely to happen anyway) but mainly because the original comments are rather more varied, forceful, and stimulating. Nevertheless, I have taken quite a few liberties per hour in the short time available for preparing this presentation. In fact, it would not surprise me if it were concluded that the number of liberties per square hour has been strongly positive but, of course, I had to make a choice with a view to this candle. Oh, by the way, it looks thin and short, but it is longer than you may have realised, for part of it is hidden in this holder, which is actually, and fittingly, a Dutch flower pot upside down, and by using my digits I can push it up so as to make it last longer. That is a clear example of digital control (1 thought we ought to have at least one demonstration at this conference). Well, here we go. Economic evaluation of computer control The question was raised whether it was necessary on the one hand, and possible on the other, to evaluate computer systems in economic terms. Insufficient availability and accuracy of data, as well as the fact that evaluation is usually required in advance of project realisation were indicated as principal difficulties.
*The author is full professor of Measurement and Control Engineering at Eindhoven University of Technology, P.O. Box 513, Eindhoven, Netherlands. 289
290
IFAC Report
Nevertheless. most participants shared the view that sound evaluation must provide the major guidance in the development of computer control systems. "'Facts" should prevail, while "faith" would mean the trust of management in the technological skills of the evaluating team, Thorough problem definition and/or appropriate process modelling may help very much in quantifying the technological and thereby also the economical effects of computer control. Regulation was prescribed as replacement (and only in DDC regarded as a one-time saving), optimisation was valued high, but at the cost of difficult models, while production control was indicated as extremely profitable but at the same time difficult to prove (intangibles !). Evaluation by switching-off the computer-control system was mentioned. The experience in one case was that no permission was given to stop the computer, which, in a way, provides at least some proof of the system's actual use. Finally, attention was drawn to the important fact that many systems tend to degrade in the course of time, if only because they are used less and less by operators. Methods of measuring operator usage were advocated, together with goal setting and correcting measures. My impression is that the discussion was overlooking a phenomenon that is quite well known in conventional process control. Already during the very first startup of a new type of plant, various instruments are found to be hardly necessary {particularly if the instrumentation was engineered by an outside contractor having little access to the relevant operating department). During subsequent operation and later start-ups, more and more instruments turn out to be relatively unimportant, one of the principal reasons being that the operators become more and more expert in running the plant and in coping with its caprices. Operators learn fast. So there is every reason to expect that they will also learn fast from computer control and hence will rely less and less upon the computer for certain critical tasks while relegating more and more other tasks to it. Even a control computer functions as a teaching tool. Consequently, its role in running the plant and its interplay with the operators merits closer attention.
Ergonomic aspects The summary of this round-table discussion mentions the following four points. 1. Ergonomic considerations should include task peculiarities, priorities, sequencing. Often, instrumentation design factors should be derived from these, not otherwise, in order to avoid leaving a mess of odd jobs to the operator, 2. Operators prefer dedicated pushbuttons over keyboards, even at the expense of manipulating whole fields of buttons. 3. The display of the 80's is colour display, computer-assisted with software modules to be selected at will. 4. Complete survey of process status remains always necessary, even in spite of such displays; cognitive aspects will become more important: operators will have to think at a higher level, more oriented towards decisions than towards control. Control engineers seem to become more and more interested in the human side of automation, as became clear in several other round-table discussions. Centralised versus distributed computer systems The notion of distributed control covers a large diversity of possible control and instrumentation schemes. There are not only many 'degrees' of distribution, but distribution can also be implemented along two different lines: according to control functions or to process functions. My personal impression is that some participants talked about (de)centralisation of hardware, whereas others were talking in terms of lower-level control functions, and still others referred to higher-level functions (coordination, etc.). Implicitly, probably all were talking about increased functional integration. It was fun to hear the remark that here we may have an area with a lack of theory instead of an abundance. There was general agreement among the participants that there is a clear trend towards decentralisation resulting in hierarchical networks of computers, microcomputers and programmable controllers, but many speakers also warned against going distributed just because it is fashionable. For each concrete application, the nature and the size of the process to be
automated and the relative advantages of centralisation and decentralisation must be carefully compared according t~, a variety of criteria, among others:
--Cost: wiring costs against hardware and software cosl,~. including those related to interfacing. --Reliability, availability, s'erviceability: when the process consists of a number of relatively independent subprocesses, a distributed control system may be less vulnerable. In other cases, a centralised system may have a higher degree of availability. ---Ease of implementation: a distributed system can be realised step by step. This is especially interesting when existing plants are modernised. -Degree oJ interaction m large multirariable processes: in case of strong interaction, centralised control systems may be the • obvious solution. There is an urgent need for normalisation of data interface standardisation methods. Notably the signal codes, but also the hardware and software of the interfacing elements ought to be standardised. According to one of the participants we also need a general theory that would help us to design optimal hierarchical control structures. The most remarkable realisation in the field of distributed control that was mentioned is a control system at Kaiser Aluminum U.S.A., in which about 100 microprocessor-based controllers are supervised by a minicomputer.
Microcomputers 1. The controversy is not between micros and minis, but between micros for general and micros for special applications. 2. To decrease software cost, the Intel approach is not very useful because the PL/M language is not really universal. 3. There is a trend to use high level languages, provided they are machine-independent and do not slow down program execution too much (program preparation on another machine). 4. It would be very useful to compare the usefulness of BASIC, PASCAL and RTL as programming tools for microsystems in Automatic Control. Standardisation efforts in process control soltware Panel and audience covered the whole spectrum from general agreement to standardisation, via scepticism as to what could be achieved, to rather negative views concerning the sheer futility of current standardisation efforts. Key 1. 2. 3. 4.
questions regarding software were: Do we need standardisation at all? Does it promote or prevent progress? What does the user expect? What does the vendor want ?
In spite of the critical questions and remarks, some kind of consensus about the need for better, common software tools seemed to exist.
Standardisation of interface systems No summary of this round-table discussion was available, perhaps because it was one of the first. Education in digital computer-control This round-table discussion went very general almost immediately--in spite of the initial attempts of the panel members to focus the attention upon the subject-matter--the main subjects of discussion being: --continuing education in control engineering in general --mastering problem solving and the fundamentals of engineering --fundamentals of sensor and measurement techniques --professional skills (reading, writing, listening, social skills) - - t h e gap between theory and practice. Three kinds of professionals were identified: 1. computer system builders 2. control engineers 3. process engineers and other noncontrol engineers,
IFAC and two categories (a) students and (b) practising engineers. At present, the needs/3f the latter two categories are not very different in the computer control field: both have to become familiar with a rapidly developing field. The necessity of education in digital systems was stressed, as were: --economics of computer control --next generation effect* --attitudes of organisations, human factors: reliability, availability, serviceability ~computer-aided design - modelling, identification, updating --'Teaching modern control theory pays off.' The latter areas of concern lead us to the round-table discussion on: Application of "Modern Control Theory" to computer control in the process industry As a result of the discussions at the preceding conference at Zurich, 1974, the initiative was taken to conduct an enquiry among industries and universities concerning the control algorismst used in actual process control applications with a view to present the findings at this conference in The Hague. The motives and the questionnaire have been published in Journal A[ 1]. The questionnaire was distributed through IFAC-channels and 61 responses were received from 18 countries, 70% of the responses coming from industry. The report on this enquiry is published in Journal A[2]. In a brief presentation the aims of the enquiry and the nature of the responses were sketched together with the principal conclusions. In connection with the gap between theory and practice--or between university and industry--the report on the questionnaire mentioned, among other items: We have to accept that in industry ---measurements will always be relatively 'poor' or too few --process parameters are too little known and constantly varying --disturbances are difficult to predict and to describe adequately --operator's acceptance of control is different from the engineer's motivation to improve it -improvements towards optimality versus cost and effort --maintaining of (advanced) control must not require highly qualified manware (I would prefer to speak in terms of brainware, though). The authors of the report I-2] also submitted the following points: It is impossible to control better than one can measure. And conversely: It is senseless to measure better than one can control anyway. Similarly, one should realise that: It is impossible to control better than people allow; and conversely that: It is senseless to motivate people better, if the process does not call for it. In view of the available time. I shall resist the temptation to dwell more extensively on the report, but I recommend you to read it, if only the epilogue. One very true remark from the audience was that 'modern control theory' cannot be blamed for not being process-control oriented, for it was stimulated and financed mainly by aerospace, hardly by the process industry. *This relates to a consequence of improved control that is probably as important as it is difficult to identify: if control innovations improve plant control, later plant designs will be adapted to exploit the benefits more fully. Just imagine what would happen if today's plants would have to be controlled using the instruments of the 1940's! "tlf still alive, the Sheik would probably not mind if those who do not care about etymology and like the similarity to logarithms, write algorithms instead.
Report
291
Another remark repeated that university work is necessarily oriented towards new things and publications, whereas engineers in industry have to get things working, and are expected to make profits, not publications. The report wonders whether process control theory is approaching an asymptote (presumably a horizontal one). During the IFAC/IFORS Conference on Dynamic Modelling and Control of National Economics. Professor M. H. Peston (Queen Mary College) made a remark about the relation between literary and quantitative economists that may apply equally well to the relation between those who apply control engineering and those who develop "modern control theory'. He said: "For a long while the literary economist lacked the courage to say the emperor* had no clothes because he could not pretend to emulate him or even understand him; and, anyway, the emperor did have some clothes--he was just rather scantily clad. What is more the emperor could reasonably demand more time. before being fully subject to the relevant test of achievement. If we date the quantitative revolution from 1945, what is now being said is that twenty-five years is long enough: it is now necessary to put up or shut up." In any case, if control theory has to have any hope of becoming really useful it must pay far more attention to realistic problem formulation and to evaluation of its practical advantages and disadvantages. But above all: to what the people on the job do really need. Mr. Chairman, We have now had five conferences in thirteen years, which reminds me of a song by a Dutch entertainer and his wife. entitled Girls of Thirteen. It goes roughly as follows: Girls being thirteen... Not very h a p p y . . . No longer children, still not yet women. Already too old for fairy tales and yet still not quite up to reality. Easy to hurt, but a wonder to behold and a marvel to watch growing up. Mr. Chairman, Girls of thirteen learn a lot about the facts of life by talking among themselves and that is exactly what we have been doing in this conference, now behind us, and in the preceding four conferences. And we have been learning a lot from each other. Purists may wonder whether the flag still covers the cargo, but the crucial question is whether a conference attracts an excellent mix of people. This one certainly did. I hope that, if there are going to be three more conferences like this one, they will be as successful as their five predecessors, for then--after 21 years-computer control will have become a grown-up science. Ladies and Gentlemen, This conference has been a strenuous one. 1 suppose that many of us look almost burnt up, like this candle, but also have ample reserves in store. So I now snuff out this candle, convinced that the flame of the conference will be kindled by you in the years to come. *i.e. the quantitative economist. References [l'l G. DUYFJES, P. J. DE JONG, and H. B. VERBRUGGEN: Questionnaire on applications of 'Modern Control Theory' to computer control in the process industry. Journal A 17(2) (May, 1976). [2] G. DUYFJES. P. J. DE JONG and H. B. VERBRUGGEN: Questionnaire on applications of 'Modern Control Theory" to computer control in the process industry. Results and comments. Issued during the Vth IFAC/IFIP Conference on Digital Computer Applications to Process Control, The Hague, 1977, 13 pp. To be published in Journal A 18(4) (November, 1977).
IFAC Report Digital Computer Applications to Process Control Summary of round-table discussions, 5th IFAC/IFIP International Conference; The Hague, The Netherlands, June 1977. O.
RADEMAKER*
Let us linger for a moment in 1964, when the first Computer Control conference took place in Stockholm. We were then talking mainly about registers, bits, instruction sets, poor computer reliability, and how to install computers. We worried about shielding and earth connections. We discussed optimal programming of a computer whose working memory was on a drum, as well as the feasibility of adc times smaller than 200 ms, and the reliability of relay multiplexers. Further, we speculated on the break-even number of ddc loops, wondering whether computer control would ever become really economically justifiable in view of the fantastically high price and unwieldy size of computers, and so on and so forth. Now look at 1977: True, of the 34 papers on methods only about three were not from Universities or Research Institutes, but there were many more papers looking back upon real applications:
Summary--A summary is given of the issues and conclusions of the round-table discussions on 'Economic evaluation of computer control', 'Ergonomic aspects', "Centralised versus distributed computer systems', 'Microcomputers', 'Standardisation efforts in process control software', 'Education in digital computer-control', and "Application of "Modern Control Theory" to Computer Control in the Process Industry'. Mr. Chairman, Ladies and Gentlemen, DURING this IFAC/IFIP conference, which is now almost behind us, one speaker remarked that automatic feedback started in the 1930's with the advent of electronics, but actually, it was already used a few thousand years ago by Ktesibios of Accra, who was the teacher of Heron of Alexandria, and probably even earlier by the Sumerians, so I hope the audience will not mind if I go back only about two hundred years for a minute. Back to the famous composer Haydn, who lived at a time when windmills and watermiils already had automatic control devices, and who wrote the Farewell Symphony (Abschiedssymphonie), also called The Candle Overture. In January 1772, his noble master Esterhazy moved his court from delightful Vienna to his lonesome winter residence and told his orchestra, conducted by Haydn, to leave their wives at home, because they were always quarrelling anyway. After a while, the members of the orchestra did not like that any longer and started to complain. Haydn then wrote this particular symphony, in which--one after the other--the musicians finished their part of the music, blew out their candles, picked up their instruments, and left, until the podium was empty. His Royal Highness understood this ungentle hint and said: 'Tomorrow we shall leave'. The memory of this story came back to me again when the organisers of the conference asked me to summarise the conclusions of the round-table discussions, adding that this kind of final meeting was bound to be somewhat disturbed by participants leaving to catch their plane, train, or wife, in time. So I decided to bring a candle myself, which made me remember another story: In a small area in our country, parents of an attractive girl had a very peculiar custom, centuries ago, which is now completely forgotten in Holland, but has been kept alive by descendants of Dutchmen who left the country centuries ago; kept alive, almost at the end of the world: If a beautiful girl goes steady and her boy friend visits her home, her father and mother go to sleep at some decent time, leaving a burning candle behind. Before the candle goes out, the boy is supposed to be out of the house (which goes to tell you how old this story is!). Now, the point is, what kind of candle do the parents leave behind ? If it is thin and short, they do not think much of the boy. If it is thick and tall, they like him. So I considered the request of the organisers, and since they did not give me a candle I have brought my own (a thin and short one).
41 application papers, 11 case histories, 10 survey papers or tutorials, and many of the old issues have virtually disappeared. Some have remained, as became clear in the eight round-table discussions the organisers asked me to talk about, using the reports by the chairmen as the main sources of information. Well, first of all, they were not round at all. Those I have witnessed were characterised by: - - a large panelled room full of listeners, - - a panel behind the table, and - - a variety of threads of discussion that could not be woven into a fabric within an hour. But I do think the round-table discussions left many participants with many questions to be discussed among themselves, and this conference, this building and above all you, the participants, offered splendid opportunities for doing that, so I have the impression that, bye and large, the round-table discussions have been quite successful in the overall framework of this Conference. The chairmen of the various round-table discussions have been invited to summarise their findings on a single sheet of paper. Most of them have done so, in a variety of ways. I considered the possibilities of functioning as a cross-compiler but decided against that idea, not because those chairmen might become cross with me (which is likely to happen anyway) but mainly because the original comments are rather more varied, forceful, and stimulating. Nevertheless, I have taken quite a few liberties per hour in the short time available for preparing this presentation. In fact, it would not surprise me if it were concluded that the number of liberties per square hour has been strongly positive but, of course, I had to make a choice with a view to this candle. Oh, by the way, it looks thin and short, but it is longer than you may have realised, for part of it is hidden in this holder, which is actually, and fittingly, a Dutch flower pot upside down, and by using my digits I can push it up so as to make it last longer. That is a clear example of digital control (1 thought we ought to have at least one demonstration at this conference). Well, here we go. Economic evaluation of computer control The question was raised whether it was necessary on the one hand, and possible on the other, to evaluate computer systems in economic terms. Insufficient availability and accuracy of data, as well as the fact that evaluation is usually required in advance of project realisation were indicated as principal difficulties.
*The author is full professor of Measurement and Control Engineering at Eindhoven University of Technology, P.O. Box 513, Eindhoven, Netherlands. 289
290
IFAC Report
Nevertheless. most participants shared the view that sound evaluation must provide the major guidance in the development of computer control systems. "'Facts" should prevail, while "faith" would mean the trust of management in the technological skills of the evaluating team, Thorough problem definition and/or appropriate process modelling may help very much in quantifying the technological and thereby also the economical effects of computer control. Regulation was prescribed as replacement (and only in DDC regarded as a one-time saving), optimisation was valued high, but at the cost of difficult models, while production control was indicated as extremely profitable but at the same time difficult to prove (intangibles !). Evaluation by switching-off the computer-control system was mentioned. The experience in one case was that no permission was given to stop the computer, which, in a way, provides at least some proof of the system's actual use. Finally, attention was drawn to the important fact that many systems tend to degrade in the course of time, if only because they are used less and less by operators. Methods of measuring operator usage were advocated, together with goal setting and correcting measures. My impression is that the discussion was overlooking a phenomenon that is quite well known in conventional process control. Already during the very first startup of a new type of plant, various instruments are found to be hardly necessary {particularly if the instrumentation was engineered by an outside contractor having little access to the relevant operating department). During subsequent operation and later start-ups, more and more instruments turn out to be relatively unimportant, one of the principal reasons being that the operators become more and more expert in running the plant and in coping with its caprices. Operators learn fast. So there is every reason to expect that they will also learn fast from computer control and hence will rely less and less upon the computer for certain critical tasks while relegating more and more other tasks to it. Even a control computer functions as a teaching tool. Consequently, its role in running the plant and its interplay with the operators merits closer attention.
Ergonomic aspects The summary of this round-table discussion mentions the following four points. 1. Ergonomic considerations should include task peculiarities, priorities, sequencing. Often, instrumentation design factors should be derived from these, not otherwise, in order to avoid leaving a mess of odd jobs to the operator, 2. Operators prefer dedicated pushbuttons over keyboards, even at the expense of manipulating whole fields of buttons. 3. The display of the 80's is colour display, computer-assisted with software modules to be selected at will. 4. Complete survey of process status remains always necessary, even in spite of such displays; cognitive aspects will become more important: operators will have to think at a higher level, more oriented towards decisions than towards control. Control engineers seem to become more and more interested in the human side of automation, as became clear in several other round-table discussions. Centralised versus distributed computer systems The notion of distributed control covers a large diversity of possible control and instrumentation schemes. There are not only many 'degrees' of distribution, but distribution can also be implemented along two different lines: according to control functions or to process functions. My personal impression is that some participants talked about (de)centralisation of hardware, whereas others were talking in terms of lower-level control functions, and still others referred to higher-level functions (coordination, etc.). Implicitly, probably all were talking about increased functional integration. It was fun to hear the remark that here we may have an area with a lack of theory instead of an abundance. There was general agreement among the participants that there is a clear trend towards decentralisation resulting in hierarchical networks of computers, microcomputers and programmable controllers, but many speakers also warned against going distributed just because it is fashionable. For each concrete application, the nature and the size of the process to be
automated and the relative advantages of centralisation and decentralisation must be carefully compared according t~, a variety of criteria, among others:
--Cost: wiring costs against hardware and software cosl,~. including those related to interfacing. --Reliability, availability, s'erviceability: when the process consists of a number of relatively independent subprocesses, a distributed control system may be less vulnerable. In other cases, a centralised system may have a higher degree of availability. ---Ease of implementation: a distributed system can be realised step by step. This is especially interesting when existing plants are modernised. -Degree oJ interaction m large multirariable processes: in case of strong interaction, centralised control systems may be the • obvious solution. There is an urgent need for normalisation of data interface standardisation methods. Notably the signal codes, but also the hardware and software of the interfacing elements ought to be standardised. According to one of the participants we also need a general theory that would help us to design optimal hierarchical control structures. The most remarkable realisation in the field of distributed control that was mentioned is a control system at Kaiser Aluminum U.S.A., in which about 100 microprocessor-based controllers are supervised by a minicomputer.
Microcomputers 1. The controversy is not between micros and minis, but between micros for general and micros for special applications. 2. To decrease software cost, the Intel approach is not very useful because the PL/M language is not really universal. 3. There is a trend to use high level languages, provided they are machine-independent and do not slow down program execution too much (program preparation on another machine). 4. It would be very useful to compare the usefulness of BASIC, PASCAL and RTL as programming tools for microsystems in Automatic Control. Standardisation efforts in process control soltware Panel and audience covered the whole spectrum from general agreement to standardisation, via scepticism as to what could be achieved, to rather negative views concerning the sheer futility of current standardisation efforts. Key 1. 2. 3. 4.
questions regarding software were: Do we need standardisation at all? Does it promote or prevent progress? What does the user expect? What does the vendor want ?
In spite of the critical questions and remarks, some kind of consensus about the need for better, common software tools seemed to exist.
Standardisation of interface systems No summary of this round-table discussion was available, perhaps because it was one of the first. Education in digital computer-control This round-table discussion went very general almost immediately--in spite of the initial attempts of the panel members to focus the attention upon the subject-matter--the main subjects of discussion being: --continuing education in control engineering in general --mastering problem solving and the fundamentals of engineering --fundamentals of sensor and measurement techniques --professional skills (reading, writing, listening, social skills) - - t h e gap between theory and practice. Three kinds of professionals were identified: 1. computer system builders 2. control engineers 3. process engineers and other noncontrol engineers,
IFAC and two categories (a) students and (b) practising engineers. At present, the needs/3f the latter two categories are not very different in the computer control field: both have to become familiar with a rapidly developing field. The necessity of education in digital systems was stressed, as were: --economics of computer control --next generation effect* --attitudes of organisations, human factors: reliability, availability, serviceability ~computer-aided design - modelling, identification, updating --'Teaching modern control theory pays off.' The latter areas of concern lead us to the round-table discussion on: Application of "Modern Control Theory" to computer control in the process industry As a result of the discussions at the preceding conference at Zurich, 1974, the initiative was taken to conduct an enquiry among industries and universities concerning the control algorismst used in actual process control applications with a view to present the findings at this conference in The Hague. The motives and the questionnaire have been published in Journal A[ 1]. The questionnaire was distributed through IFAC-channels and 61 responses were received from 18 countries, 70% of the responses coming from industry. The report on this enquiry is published in Journal A[2]. In a brief presentation the aims of the enquiry and the nature of the responses were sketched together with the principal conclusions. In connection with the gap between theory and practice--or between university and industry--the report on the questionnaire mentioned, among other items: We have to accept that in industry ---measurements will always be relatively 'poor' or too few --process parameters are too little known and constantly varying --disturbances are difficult to predict and to describe adequately --operator's acceptance of control is different from the engineer's motivation to improve it -improvements towards optimality versus cost and effort --maintaining of (advanced) control must not require highly qualified manware (I would prefer to speak in terms of brainware, though). The authors of the report I-2] also submitted the following points: It is impossible to control better than one can measure. And conversely: It is senseless to measure better than one can control anyway. Similarly, one should realise that: It is impossible to control better than people allow; and conversely that: It is senseless to motivate people better, if the process does not call for it. In view of the available time. I shall resist the temptation to dwell more extensively on the report, but I recommend you to read it, if only the epilogue. One very true remark from the audience was that 'modern control theory' cannot be blamed for not being process-control oriented, for it was stimulated and financed mainly by aerospace, hardly by the process industry. *This relates to a consequence of improved control that is probably as important as it is difficult to identify: if control innovations improve plant control, later plant designs will be adapted to exploit the benefits more fully. Just imagine what would happen if today's plants would have to be controlled using the instruments of the 1940's! "tlf still alive, the Sheik would probably not mind if those who do not care about etymology and like the similarity to logarithms, write algorithms instead.
Report
291
Another remark repeated that university work is necessarily oriented towards new things and publications, whereas engineers in industry have to get things working, and are expected to make profits, not publications. The report wonders whether process control theory is approaching an asymptote (presumably a horizontal one). During the IFAC/IFORS Conference on Dynamic Modelling and Control of National Economics. Professor M. H. Peston (Queen Mary College) made a remark about the relation between literary and quantitative economists that may apply equally well to the relation between those who apply control engineering and those who develop "modern control theory'. He said: "For a long while the literary economist lacked the courage to say the emperor* had no clothes because he could not pretend to emulate him or even understand him; and, anyway, the emperor did have some clothes--he was just rather scantily clad. What is more the emperor could reasonably demand more time. before being fully subject to the relevant test of achievement. If we date the quantitative revolution from 1945, what is now being said is that twenty-five years is long enough: it is now necessary to put up or shut up." In any case, if control theory has to have any hope of becoming really useful it must pay far more attention to realistic problem formulation and to evaluation of its practical advantages and disadvantages. But above all: to what the people on the job do really need. Mr. Chairman, We have now had five conferences in thirteen years, which reminds me of a song by a Dutch entertainer and his wife. entitled Girls of Thirteen. It goes roughly as follows: Girls being thirteen... Not very h a p p y . . . No longer children, still not yet women. Already too old for fairy tales and yet still not quite up to reality. Easy to hurt, but a wonder to behold and a marvel to watch growing up. Mr. Chairman, Girls of thirteen learn a lot about the facts of life by talking among themselves and that is exactly what we have been doing in this conference, now behind us, and in the preceding four conferences. And we have been learning a lot from each other. Purists may wonder whether the flag still covers the cargo, but the crucial question is whether a conference attracts an excellent mix of people. This one certainly did. I hope that, if there are going to be three more conferences like this one, they will be as successful as their five predecessors, for then--after 21 years-computer control will have become a grown-up science. Ladies and Gentlemen, This conference has been a strenuous one. 1 suppose that many of us look almost burnt up, like this candle, but also have ample reserves in store. So I now snuff out this candle, convinced that the flame of the conference will be kindled by you in the years to come. *i.e. the quantitative economist. References [l'l G. DUYFJES, P. J. DE JONG, and H. B. VERBRUGGEN: Questionnaire on applications of 'Modern Control Theory' to computer control in the process industry. Journal A 17(2) (May, 1976). [2] G. DUYFJES. P. J. DE JONG and H. B. VERBRUGGEN: Questionnaire on applications of 'Modern Control Theory" to computer control in the process industry. Results and comments. Issued during the Vth IFAC/IFIP Conference on Digital Computer Applications to Process Control, The Hague, 1977, 13 pp. To be published in Journal A 18(4) (November, 1977).