This action might not be possible to undo. Are you sure you want to continue?

**PID-deadtime control of distributed processes
**

F.G. Shinskey*

Process Control Consultant, 260 Whiteface Rd., North Sandwich, NH 03259, USA Received 6 April 2001; accepted 6 April 2001

Abstract While model-based controllers have been used successfully to control paper machines and other processes dominated by time delay, matching the model to the process gives poor load regulation over lag-dominant processes. An important class of lagdominant processes including heat exchangers and distillation columns consists of distributed lags. A PID controller having timedelay compensation, while functionally similar to a model-based controller, is a much better load regulator, and twice as eﬀective as a conventional PID controller on these processes. It is applied in this paper to regulate steam superheat temperature. r 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Adaptive control; Distillation columns; Distributed-parameter systems; Heat exchangers; Load regulation; Model-based control; PID control; Steam plant; Temperature control; Time delay

1. The PID-deadtime controller The insertion of a time delay into the integral feedback circuit of a PID controller produces a transfer function similar to a Smith predictor or an Internal Model Controller. Its internal conﬁguration is shown in Fig. 1. Integration is accomplished by positive feedback of controller output m through time delay td and integral lag I. The controlled variable c passes through a ﬁltered derivative block of time constant D and ﬁlter aD; but set point r does not. An additional ﬁlter tf is located so as to be in both the process and integral feedback loops. The transfer function of m responding to c is mðsÞ Kc ð1 þ DsÞð1 þ IsÞ ¼ ; cðsÞ ð1 þ aDsÞ½ð1 þ IsÞð1 þ tf sÞ À eÀtd s ð1Þ

against integral windup and prepared to resume control when the feedback connection is restored. An internal model controller (IMC) is shown in Fig. 2 for comparison. Its transfer function is mðsÞ 1 ¼ ; cðsÞ Km ðgm * =gf À gm Þ

ð2Þ

where Km is the steady-state gain and gm is the dynamicgain vector of both the process and its model; gm* is the invertible part of gm, and gf is a ﬁlter. For the simplest case where the process is a ﬁrst-order lag with deadtime, the IMC function becomes mðsÞ 1 þ t1 s ¼ ; cðsÞ Km ð1 þ tf s À eÀtd s Þ

where s is the Laplace operator and Kc is the proportional gain of the controller. This PID form is known as the ‘‘interacting’’ or series controller; time-delay compensation can also be added to the ‘‘noninteracting’’ or parallel controller. The connection from the controller output to the time-delay compensator is shown dashed because it can be broken, allowing another signal to be substituted for the output signal. In this way, the controller can be protected

*Tel/fax: +1-603-284-6404. E-mail address: shinskey@msn.com (F.G. Shinskey).

ð3Þ

where t1 is the process lag, tf the time constant of a ﬁrstorder ﬁlter, and td the process deadtime. If the integral time of the PID controller with deadtime were set to zero as well as its derivative ﬁlter, the transfer function of Eq. (1) would reduce to that of Eq. (3). The PID controller with deadtime is then seen to be similar to a second-order IMC controller. However, this paper is intended to illustrate their diﬀerences more than their similarities.

0967-0661/01/$ - see front matter r 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 9 6 7 - 0 6 6 1 ( 0 1 ) 0 0 0 6 3 - 6

the ratio of time constant in the path of the load input to the deadtime is 5–10. the PID controller with timedelay compensation is evaluated as a regulator for distributed-lag processes. The ratio of the integrated error IE for the two responses is IEbest 1 À eÀtd =tq ¼ .G. a response identiﬁed as the ‘‘best’’ achievable. 3 illustrates the case where Kq=Km. Time-delay compensation is added to the integral feedback loop of the PID controller. Their diﬀerenceFgiven a matched modelFis the process load q after passing through its steady-state and dynamic gains Kqgq. 2. Fig. If . which is subtracted from its real value. after the deadtime in the load path elapses the controlled variable will begin to deviate from set point. 1. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 Fig. Distributed-lag processes Multiple-stage chemical processes consist of a series of connected ﬁrst-order lags of similar time constant. The IMC is designed to produce a desired set-point response: cðsÞ g ðsÞ ¼ gf ðsÞ m rðsÞ gm * ðsÞ ð4Þ Given only ﬁrst-order dynamics. Internal model control conﬁguration. In parallel with the process is its model. approaching the ‘‘best’’ load response without compromising set-point response. The problem is illustrated in reference to the block diagram of IMC in Fig. where tf is the time constant of a ﬁrst-order ﬁlter in the IMC controller. 3 is only twice the deadtime. includes the load dynamics as a dominant term: cðsÞ gm ðsÞ ¼ Kq gq ðsÞgf ðsÞ À1 : ð5Þ qðsÞ gm * ðsÞ Following a step change in load. This produces an estimated value of c designated as ce. Also shown is a recovery trajectory which returns to set point at the end of the deadtime. The advantage of the PIDtd controller lies in that it is tunable. 3.1178 F. the curves in Fig. also identiﬁed as Kmgm. then returning to set point on another exponential trajectory. so as to produce a particular set-point response. assuming a perfect match. 2. At that time. 3. Fig. It is obtained by causing m to overshoot the step in q.5–10. c will follow an exponential trajectory until the deadtime in gm elapses. 2. however. And when tuned to minimize integrated absolute error (IAE) or some similar objective function. it outperforms both PID and model-based controllers. 1996) has shown that this results in poor load regulation for processes that are lag-dominant.5. Fig. The load response. giving integrated-error ratios of 5. For lag-dominant processes. Even without a ﬁlter. In this paper.5 times the ‘‘best’’. 3 show an integrated error for IMC that is 2. IMC gives unacceptable load response. controller output m will step an amount equal to the step in q multiplied by Kq/Km. both governed by gq. Limitations of IMC While model-based controllers are intended to match the process parameters as closely as possible. For most ﬂuid processes where temperature and composition are controlled. much like a conventional PID controller. The time lag in the load path for the curve in Fig. the author (Shinskey. IEIMC 1 þ tf =td ð6Þ which includes only the process deadtime and the ﬁlter. with the overshoot sustained for one deadtime.

causing a secondary change to the composition of the overhead vapor. Each change in either the ﬂow or composition of either the vapor or liquid thus propagates both upward and downward. but both its equivalent time delay and ﬁrst-order lag change with ﬂow. In a distillation column. there is almost no diﬀerence in the shape of the response curve when n=100Fit does not change signiﬁcantly for values of n>20. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 1179 information can ﬂow in both directions through such a process. The factor (n2+n)/2 is equal to the where Q is the rate of heat ﬂow and T1 and T2 are the respective inlet and outlet temperatures of the liquid. forming a ‘‘ladder network’’. But most columns contain 20 or more trays.2% response occurs at a normalized value of 1. for example. heat exchangers generally operate on a once-through basis. However. By contrast. An electrical analog of such a multistage process is a series of identical resistors with identical capacitors connected across the line at each junction.1. their dynamics tend to be unaﬀected by production rate. the 20-lag model is satisfactory for simulating much higherorder systems. ð8Þ where ti is the value of the individual time constant and n is their number. ultimately aﬀecting the composition of both overhead and bottom products in a series of steps. which subsequently aﬀects the composition of the vapor leaving the next tray. Consider the following steady-state model of heat transfer between steam ﬂowing at rate W and having a latent heat of H. 3. The steady-state gain of exit temperature in response to steam ﬂow is seen to vary inversely with liquid ﬂow: Kp ¼ dT2 H : ¼ dW FC ð9Þ Fig. Open-loop response Fig. and a liquid ﬂowing at rate F and having a speciﬁc heat of C: Q ¼ WH ¼ FCðT2 À T1 Þ. making it easier to control. . a step change in reﬂux ﬂow entering the top will aﬀects the composition of the overhead vapor almost immediately. but also with some initial time delay. This new vapor composition then shifts the equilibrium on the top tray. As a result. Another example of a distributed lag with constant dynamics is a stirred tank.2. Perhaps the most common distributed process is the heat exchanger: heat is transferred across a broad area having a distributed temperature gradient and distributed heat capacity as well.00. 4. This squared relationship can give a 100-tray distillation column a four-times slower response than a 50-tray column. The step response of 20 interacting lags. control-loop stability must also be provided at low ﬂow rates. The steady-state response of extrinsic variables such as temperature and composition also depends on ﬂow rates. Given that an exchanger has a ﬁxed heat capacity. Variable parameters Distillation columns tend to operate with internal circulation rates of liquid and vapor which are much higher than their throughput. the exit temperature responds to changes in steam ﬂow faster and with less gain. The time scale is normalized so that 63. The shape of the response curve does not change. neither ﬂuid being recirculated. Interestingly. ð7Þ sum of the numbers from 1 to n. This network has also been used to simulate the response of electrical and pneumatic transmission lines. the variable ﬂow rate of the ﬂuids causes residence time St to vary inversely with ﬂow. these lags are considered to be ‘‘interacting’’. Note that it is dominated by the exponential response of an equivalent ﬁrst-order lag. 4 shows a step response of a network of 20 equal interacting lags.F.G. the shape does not change signiﬁcantly above 20 lags. where the response is slower and the gain is higher. while keeping the same ratio between them. 3. As a consequence. But it also changes the ﬂow and composition of the liquid leaving the top tray. especially when there are many trays. providing that its internal circulation rate is reasonably constant and greater than throughput. the individual steps can sometimes be seen in the response of product compositions to step disturbances. Distillation columns containing packing rather than discrete trays behave as distributed lags. consisting of an inﬁnite number of inﬁnitesimal lags. If the stages are few in number. The time at this point is St ¼ ti ðn2 þ nÞ=2. whose capacity smoothens the composition response curves into long exponential lags. As liquid ﬂow increases. The highest-order interacting system is the distributed lag.

5.3. (The proportional gain Kc=100/P.54 0. As a consequence. but by using a controller whose structure does not match the process.4.) Integrated error can be related to operating cost in most product-quality loops.5 for the ﬁrst-order-plus-delay process and 2. Control-loop robustness Fig. the PIDtd controller seems particularly well-suited to regulating distributed processesFover seven times as eﬀective as PI control on the basis of integrated error. 4. In other words. Proceeding from a PID controller to one with time-delay compensation is even more striking: the integrated error is reduced by a factor of 1.6 I=St 0.10 0. So the concept of matching the controller to the process must be rejected on two counts: a higher performance can be achieved not only by tuning the controller instead of matching its parameters to the process. Tuning vs. Table 1 lists their mode settings in terms of the process parameters Kp and St: Proportional band P is expressed in percent. The optimum controller tuning is somewhat diﬀerent as well. the load response would be much poorer. and steam superheat. the intent being to match the process delay with one in the control system. For example. While elimination of tuning may be desirable.89. except for that of the ﬁlter. and interacting PIDtd : All controllers were tuned to minimize integrated absolute error (IAE). calculated as IE P ¼ ðI þ td Þ: Dm 100 ð10Þ to avoid lifting safety valves. due to the dominant lag in the load path.067 D=St F 0. either. What is proved instead is that the PIDtd controller has a high performance capable of maximizing load response on a variety of processesFits structure does not depend on the nature of the process being controlled. along with the elimination of tuning. although its tuning does. interacting PID. where extreme temperatures shorten the life of metal heat-transfer tubes. Examples are the control of boiler pressure. 3. Closed-loop response Modeling a distributed-lag process as ﬁrst-order plus time delay is not very satisfactory in predicting its closed-loop response. In some applications. all three controllers were tuned to minimize IAE.) The last column in the table is the integrated error (IE) per unit change in controller output m required to meet the load change. the peak deviation may be as important.G. the set-point response of this controller is also excellent with the tuning that is optimal for load response. the high performance of the PIDtd controller on the distributed process refutes the notion that model-based control is superior to other methods. Fig.1180 F. In the distributed process. Improved set-point response was promised. the reduction achieved by the same change in controllers is 2.16 td =St F F 0. where it represents valuable ingredients given away or excessive energy consumption used to assure that speciﬁcations are met. matching Model-based control was developed primarily for processes having a pronounced time delay. Step load responses for a distributed lag. But for a distributed process.11 (KpSt) 0. The distributed lag is much more responsive to derivative action in the controller. (Although not a part of this discussion. this exercise shows that it is still required if load rejection is to be optimized. So the progression of proportional-band settings in the table relates to the peak deviations shown in Fig. yet a controller having a time delay is more eﬀective regulating it than those that do not. Furthermore.5 for the distributed process. if the PIDtd controller were replaced with a model-based controller matching the process faithfully. 5. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 Table 1 Controller settings and integrated error for optimum load regulation of distributed processes Controller PI PID PIDtd P/Kp 20 15 6. there is no true time delay. however. the step load response of a ﬁrst-order process with deadtime can be reduced in integrated error by a factor of 2. The size of the peak deviation following a load change is directly proportional to the proportional-band setting.25 by changing from PI to interacting PID control.25 0. Robustness has been deﬁned as the minimum change in given process parameters which brings the loop to the .015 (KpSt) 3.038 (KpSt) 0. especially for a lagdominant process.16 IE/Dm 0. 5 compares the step load response of a distributed process (simulated by 20 interacting lags) using three diﬀerent controllers: PI.

4 is analyzed in Fig. and the time along that slope is the estimated time constant. in the case of a heat exchanger. With the PIDtd controller in the loop. Their method. projected if necessary. the lower the robustness of the loop. However. Time-delay compensation. The size of the step disturbance Dm is marked against the steepest slope. Process identiﬁcation If the PID controller with time-delay compensation is to be used to control a heat exchanger.G. For the PID loop. but together. 6 shows that the estimated time delay tde lies between the initiation of the step at time zero. the increase in performance which it brings. longer derivative time). in that the dominant lag of a distillation column could be more than an hour. A self-regulating process really requires three parameters for identiﬁcation: time delay. (The converse is not necessarily true. and more robustness when moved in the other (lower gain.1. as if the steadystate gain were unity.F. they replaced these two parameters with one. in that it is possible for a controller to exhibit low performance and low robustness at the same time. having a consistently low ratio of time delay . integral. In that the distributed lags are in fact lag dominant. can cause instability when pushed too far in either direction. and steady-state gain.1. or rose to 112% of the originalFthe last being the narrowest of the margins observed. A step change is manually introduced into the controller output from a steady state. Evaluating the reaction curve Ziegler and Nichols apparently did not want to wait to achieve complete response of lag-dominant processes. For this application. time constant. Application to the distributed lag One of the concerns when using the Ziegler and Nichols tuning rules is their limited scope. 5.). 4. comes with the requirement of very precise tuning. 6 shows how to estimate the dominant time constant t1e . 6. The step-response curve for the distributed lag from Fig. other disturbances could arise. As a general rule. and derivative mode settings provide tighter control when moved in one direction (higher gain. 5. inversely with process ﬂow. Estimating the time delay and time constant using an openloop step response. Robustness considerations with heat exchangers There are only two parameters required to specify the distributed lagFKp and StFand as indicated in Table 1. The earliest and easiest identiﬁcation method to administer is the step test used by Ziegler and Nichols (1942). Fig. then. Fig. used only two features of the curve upon which to base their controller settings: the estimated time delay and the steepest slope of the curve. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 1181 limit of stability. an estimate of the gain requires a return to the steady state. etc. Fig. the limit of stability was reached when the ﬂow fell to 63% of the value where the controller was optimally tuned. and the intersection of the steepest slope with the baseline. However. faster integration. therefore. 6. and so terminated their test after observing the slope of the reaction curve pass its steepest point.) Adjustments to proportional. In simulations of the PI loop. which may require too much time. as identiﬁed by an undamped oscillation. Reasoning that the steepest slope was a function of both the gain and the time constant. the higher the controller performance. and may not be reached at all. This has merit.2. the robustness of the PIDtd controller is only 12%. They work best for step load changes applied to lag-dominant processes. and the resulting response in the controlled variable recordedFthe ‘‘reaction curve’’. to tune the controller. compared to 31% for the PID controller and 37% for the PI controller. In an alternative to calculating the slope. In the meantime. robustness properties for the three loops described in Table 1 and Fig. 5. which has been the experience of some engineers attempting to control pH or steam temperature using a Smith predictor. 3 need to be evaluated as these two process parameters change together with ﬂow. achieved either through tighter tuning or the addition of more modes. As a consequence. instability was reached at 69% of the original ﬂow. instability was reached when the ﬂow fell to 79% of the original. the dynamics of the process need careful identiﬁcation across the entire operating range. however. they do not vary independently. Therefore. and ﬁve time constants are required for 99% complete response.

(9).5 : 1 range.1 2. which the distributed lag does not. 6.5 3. This means adapting all the mode settingsFnot only the gainFbased on the observed response of the process. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 to time constant. Plant identiﬁcation The PIDtd controller was applied to the superheaters of a 500 MW power boiler in Ontario. its value comes out to be 0. The results of those tests appear in Table 2. The ﬁrst correlation above has no other function than to identify the steady-state gain. therefore. (10) gives a result of 0.4 . 6.2. the process parameters cannot be expected to vary with steam ﬂow following a simple formula such as Eq. Over the range tested. After obtaining the estimated process The distributed-lag process has.89 9.G. other Table 2 Parameters estimated from step tests Load (MW) Steam ﬂow (%) tde (min) t1e (min) Kp 100 200 300 400 530 20 35 50 66 89 3. Because of the robustness limitations of the controller. Controller performance on this process can only be estimated relative to other controllers.1182 F. (9). generated load varied by 5.3 : 1. Controlling a superheater One of the most diﬃcult loops in a modern power plant is that of controlling superheated steam temperature. However steam ﬂow is the variable used to index the controller settings. There is even a slight reversal in the estimate of Kp. To demonstrate.25 24. As steam ﬂow is increased to satisfy increasing demand for electrical power. Based on Eq. Therefore.5 2. a value that is clearly unrealistic.3 1. probably caused by a change in another operating variable such as ﬂue-gas recirculation. 3. with the time delay and time constant estimated as in Fig. and the mode settings would need to be gain-scheduled as a function of steam ﬂow. 6. While the process is truly a distributed lag. where a 12% ﬂow increase moved the loop from optimum performance to undamped cycling. the process parameters had to be determined precisely across the full load range.5 3. half as much. as it changes with load. Canada. In this context.5 2.98 16. as seen in the second correlation above. St represents the sum of the equivalent ﬁrst-order lag and delay. but its actual variation was only 2. Excessive temperatures.5 3. 1.3 3. and especially temperature variations. an equivalent ﬁrst-order time constant of 6 times its delay. it was mentioned that a ﬁrst-order lag with deadtime is not a particularly useful simulation of a distributed process. Steam temperature was to be controlled at 538751C. but is controlled by spraying attemperating water between upstream sections. Therefore. and its parameters generally vary inversely with ﬂow. The eﬃciency of the generating station depends strongly on maximizing this temperature. just as shown in Fig.1.0 4. but within very narrow constraints. Earlier. Kp would be expected to vary as much. Yet even these moderated variations are huge compared to the narrow robustness limits demonstrated by the PIDtd controller in simulated closed-loop testing.5 St (min) 5.90 17. and that changed over a 4. as shown for a ﬁrst-order process in Fig. To make use of the tuning rules given in Table 1 requires conversion from the parameters of tde and t1e to Kp and St: Given that all distributed processes have the same shape for their reaction curve.25 7. a direct correlation is possible: Kp ¼ 7:5tde =tle St ¼ 7:0tde : ð11Þ factors are also adjusted. 6.8 4.0150 Kp St: This would indicate a controller performance of 0. showing a variation of 3. their method is reasonably eﬀective here.3 : 1. The ‘‘best’’ integrated error estimate really only applies to processes having deadtime.0219 Kp St: Then estimating the integrated error produced under PIDtd control by inserting its settings from Table 1 into Eq. the relationships in the context of the boiler are somewhat more complex. The dynamic parameter St followed steam ﬂow more closely. cause stress and distortion which can signiﬁcantly shorten the life of a superheater. conducted at load levels from 100 through 530 MW.0219/ 0.0150 or 146%. Implementation The temperature controller was implemented by inserting a time-delay compensator in the integral feedback loop of an interacting PID controller. including the tilt of the burners and the recirculation of ﬂue gas.35 1. Open-loop step tests were. the use of the parameters estimated from the reaction curve is limited to determining optimum controller settings. Allowable temperature is limited by the ability of the steel-alloy heat-transfer tubing to retain its strength.3 : 1. The steam temperature is measured after leaving the last of a series of heat-transfer sections. but must be determined on-line by testing at several load levels.05 5. The controller would have to remain tuned very precisely to deliver the expected performance improvement over PID control.5 2. if the deadtime and time constant parameters estimated from the stepresponse curve are used to calculate the ‘‘best’’ integrated error for a unit load change.

cycling can develop if the deadtime setting is either too long or too short. (1996).G. Control of steam temperature proved to be tighter than had ever been achieved for that loop previously. Acknowledgements The author gratefully acknowledges the work of * Sigifredo Nino of Foxboro Canada. Ziegler. Transactions of the ASME. 130–132). who conducted the plant tests and implemented the control system so successfully. Adding a fourth parameterF deadtimeFcomplicates the procedure much more. In this way. the optimum settings found for each of the load levels were programmed into a gain-scheduler.F. References Shinskey. the controller has been providing satisfactory performance for over a year. Shinskey / Control Engineering Practice 9 (2001) 1177–1183 1183 parameters in Table 2.) (pp. Process control systems (4th ed. with similar results. G. the controller was then tuned following the rules given in Table 1. 759–768. At the time of writing this paper. no ﬁne tuning was required at any loadFthe rules which had been developed by simulation of a process consisting of 20 interacting lags proved to be entirely satisfactory. (1942). F. and it is not always clear in which direction to adjust it when attempting to stabilize a cycling loop. After achieving stable. Remarkably. Optimum settings for automatic controllers. N. It is common experience to ﬁnd even the derivative setting of PID controllers left at zero by operators and technicians because it complicates the tuning procedure. the controller was retuned at each of the other levels. Finally. responsive control at one load level with these settings. the controller would remain optimally tuned at all loads. in that the PIDtd controller is not easy to tune on-line. . New York: McGraw-Hill. J. In addition.. This is an important consideration. which interpolated between the values obtained by testing. & Nichols. B. G.

- Y4263B
- FT 160MT
- Analysis of a Melting Model for an Extruder With Reciprocation
- [Goncalvès2005] DiscretisationEDP EDO
- Xlpe Pvc 600v Sere3bf
- Ductos de Barra Zucchini
- BusducTaijan
- Legrand MCCB
- Presentation Robust Control
- NSPB-MVLV_catalog[V7]
- [Balaji2010]Passivity Based Control of Reaction Diffusion Systems
- Alterego 01
- 1982,Leja,Surface Chemistry of Froth Flotation
- Legrand MCCB
- [Ortega2001]Putting Energy Back in Control
- [Zheng2012] an Energy Saving Factory-Validated
- Chou Lak
- Extrusion 2012
- Medidas de Temp en Devanados
- Theory of Design of Control Systems
- Developments in Monitoring and Control Of
- Neuro PID Tracking Control Air Discharger
- 43668093 Apuntes de Huesca de Mecanica Del Medio Continuo

Sign up to vote on this title

UsefulNot useful- Basic Understanding of PID Controllers
- CE 434 Exam 3 Solution
- Chapter 34 Advanced Control for the Plant Floor 2010 Instrumentation Reference Book Fourth Edition
- Process Exp 4
- Sub 157255
- Internal Model Control. 4. P I D Controller Design-Rivera 86
- 2628-7651-1-PB.pdf
- Fz 3410961102
- Motor_Fuzzy_ASEE2011_Meah_Nguyen_Martin_Vaisakh_Final.pdf
- Metodika Nastavovania PID-IMC2
- 0724_FrA04.6
- Stability B L
- Introduction to Control Systems
- Temperature Control System
- Fuzzy Power EL Boost Conv
- Process Control
- Wei_He
- Inverted Pendulum controle
- MathWorks India - PID Control Design Made Easy
- Answer Key-CHE 4353-Exam 3-Fall 14
- Boiler Comb
- 4 Governor Tutorial
- 3244Plantweb TT
- ion & Control Eng'g
- Fundamentals_of_Flow_characterization_d350407x012
- Valve Characteristics
- Lab 10 Mechanical Motion
- heat
- rock bed strge.pdf
- (1) Heat, Expansion, Gas Laws, Transfer of Heat.doc
- Pid Deadtime Control of Distributive Process