¡Descarga The Role of Process Engineers in Process Control: Bridging Design and Control y más Monografías, Ensayos en PDF de Control de Procesos solo en Docsity!
ith the ever growing complexity of today’s processes and the increas- ing demands on process perfor- mance, the role of process engi- neers in the design and analysis of process con- trol strategies has to change. This is true both at the initial design stage and in the analysis of control problems in an operating plant. Two factors are driving the change:
1. Today’s emphasis on shortening the de- sign cycle is placing a lot of stress on the tradi- tional approaches to developing the process and instrumentation diagram (P&ID). 2. Corporate rightsizing has reduced, if not eliminated, the cadre of in-house control engi- neers. Those with the most experience were often the first to qualify for early retirement. Some were immediately hired as consultants, but this pool of expertise becomes diluted with time. These two factors should force changes that are long overdue. The integration of process design and pro- cess control was deemed to be desirable as long ago as the 1960s. Process engineers and control engineers tend to mix about as well as oil and water, however. They remain mixed as long as you agitate; but as soon as the agitation stops, they quickly separate. By now, most companies have progressed
to some form of digital controls. The available technology includes distributed control sys- tems (DCSs), programmable logic controllers (PLCs), personal-computer-based controls, and microprocessor-based single-loop con- trollers. The selection of the type of technolo- gy and the specific supplier frequently has been a trying exercise akin to a holy war. Most firms have no interest in repeating this endeavor. Usually, the decisions are based on issues that have little to do with process control. The two features that receive the most attention are the operator interface and the alarm package. (There is no shortage of instant experts on these two topics.) Naturally, the suppliers re- spond to this attention, and now are introduc- ing features such as three-dimensional graph- ics. What this has to do with process control is obscure. In reality, these all are system issues, not process issues. Today, the various approaches and even the available products are largely equivalent. If the job can be done with one, it can be done with any of them. So, why do we need to spend time on the system issues? The time has come to re- emphasize the process issues. This involves two activities: Design. The key is for the P&ID to truly re- flect the nature of the process. Deficiencies in
Process Engineers:
Take Control!
Chemical Engineering Progress August 2000 19
W
P ROCESS C ONTROL
Process control is too important
to be left to control engineers.
And, dealing with control issues is
easier than you might think.
Cecil L. Smith, Cecil L. Smith, Inc.
20 August 2000 Chemical Engineering Progress
P ROCESS C ONTROL
the P&ID cannot be overcome by ap- plying automated tuning, expert sys- tems, or the like. Evolution. Even with the best of efforts, the initial P&ID will not be perfect. In fact, there are advantages to starting simple and then enhancing the diagram to address the real defi- ciencies, not the “what if…” and “it would be nice to…” issues. The con- trol strategy also needs to evolve with improved understanding of the pro- cess, with new measurement technol- ogy, etc. Understanding the process is truly the key to success in both of these endeavors. Who best under- stands the process? The process en- gineers (if not, we are really in trou- ble). It then follows that the process engineers are in the best position to address the design and evolution of the process controls.
The practice
of process control
The reaction of some process engi- neers to the notion that they should address the control issues may resem- ble cardiac arrest. This response seems most pronounced among those who, while at university, took the process control course, which usually is a course in mathematics. Process engineers do need an introduction to process control — but not this type of introduction. Traditional process-control cours- es place great emphasis on topics such as Laplace transforms. Manuals from control system suppliers occa- sionally express relationships in the Laplace domain, but they do so for notational convenience. What a prac- titioner needs to know about Laplace transforms can be taught in 15 min- utes. Courses that spend more time than this on Laplace transforms are courses in mathematics. The keys to process control are the controlled variables, the manipulated variables, the disturbances, and the dependent variables. These are best understood from Figure 1. The values of the controlled vari-
ables and the dependent variables de- pend upon the values of the manipu- lated variables and the disturbances. The controlled variables must be maintained at or near a target, al- though occasionally a range of values for one or more controlled variables is acceptable. Although targets are not provided for the dependent vari- ables, constraints often must be con- sidered. The values of the manipulat- ed variables are at the discretion of the control system, but are possibly subject to constraints. The values of the disturbances are determined by factors other than the control system. To maintain each controlled vari- able at its target, the number of ma- nipulated variables must at least equal the number of controlled vari- ables (a “square” system). In most process applications, there are more manipulated variables than con- trolled variables (a “fat” system), and this creates opportunities for optimization. Be careful with the terms CV and MV. They could mean controlled variable and manipulated variable. But, alternative terms for controlled variable are process variable, PV, and measured variable, which often is designated as MV. To further confuse things, the controller output some- times is called the CV. (In this article, we’ll limit abbreviations to the famil- iar P&ID and PID (proportional-inte- gral-derivative) controller.)
Control configurations are domi- nated by single-loop control, where a controlled variable is maintained at its target by using only one manipulated variable. The control logic is largely the traditional PID control equation. Although the “shrink wrap” has been significantly enhanced through the progression from pneumatic to elec- tronic to digital technology, the PID control equation in today’s latest-and- greatest (whatever you deem that to be) is the same as in the large-case pneumatic controllers from the 1940s. The improved shrink wrap offers some opportunities, however, one of which will be discussed subsequently. Single-loop PID control is ade- quate for a very high percentage (90% or more) of the controlled vari- ables in most plants. Engineers pay homage to the “keep it simple” prin- ciple, but, sometimes, much like politicians honor a “no new taxes” pledge. With all of the fancy features of modern control systems, the temp- tation to play with the toys is irre- sistible. The result is “creeping ele- gance,” the most obnoxious effect of which is configuring an excessive number of process alarms. The nature of the control strategy primarily depends upon the relation- ships between the controlled vari- ables and the manipulated variables. There are two aspects of these rela- tionships, both of which are deter- mined by the process:
Manipulated Variables
Multivariable Process
Dependent Variables
Disturbances
Controlled Variables
Figure 1. The essence of process control.
inappropriate P&IDs will be the least of our problems). Based upon this un- derstanding, proceed as follows:
- List the controlled variables. Process variable transmitters are required for each of these, al- though, occasionally, a controlled variable will be computed from other measurements.
- List the manipulated variables. A final control element is required for each. From an instrumentation per- spective, the manipulated variable is the signal from the control system to the final control element. But, for control valves and pumps with vari- able speed drives, process engineers can consider the manipulated variable to be the flow. Where flow measure- ments are feasible, flow controllers should be installed. From the instru- mentation perspective, the manipulat- ed variable is the set point of the flow controller, but, from a process per- spective, the manipulated variable is the flow.
- Assess the degree of influence of each of the manipulated variables on each of the controlled variables. For the initial cut, only a qualitative assessment is required. Consider dy- namic aspects, especially dead time, as well as steady-state aspects. For each manipulated variable, identify all controlled variables that it signifi- cantly affects. Conversely, for each controlled variable, determine all ma- nipulated variables that significantly affect it. A matrix arrangement often is convenient.
- Identify the simple loops. The best situation is when a controlled variable is significantly affected by only one manipulated variable; here, use single-loop PID control (except for some special cases that we’ll dis- cuss later). When the controlled vari- able is affected far more rapidly by one manipulated variable than by any other, single-loop PID control usually proves adequate. For example, dy- namic considerations dictate that the bottoms composition for a distillation column must be controlled by manip- ulating either the bottoms flow or the
boilup, not by the overhead flow or the reflux.
- Check for dynamic separation. Interaction is a problem only for loops with approximately the same response speed. If one loop is much faster than another, the fast loop must be tuned first, and then the slow loop can be tuned. The slow loop may not function when the fast loop is in man- ual, but this is best addressed by im- proving the reliability of the fast loop.
- Modify the menu of manipulat- ed variables. You can introduce ra- tios, sums, differences, function gen- erators, etc., to replace one manipu- lated variable by another. For exam- ple, the stack oxygen from a combus- tion process depends upon the fuel flow and the air flow. The stack oxy- gen, however, really is a function of the air-to-fuel (or fuel-to-air) ratio. Instead of considering the manipulat- ed variables to be the air flow and the fuel flow, consider the manipulated variables to be the fuel flow and the air-to-fuel ratio (or the air flow and the fuel-to-air ratio). Applying this approach to a process requires sub- stantial insight into the behavior of the process. There is no methodology for determining where ratios, sums, etc., would be beneficial. But, cre- ativity here can be very rewarding.
- Address the multivariable is- sues. For the remaining controlled variables, the potential for serious in- teraction exists; this is not necessarily the case, however. Multivariable is- sues can be approached in two ways: Process understanding. If you un- derstand the process well enough, you should be able to sort out the in- teraction issues. Most control engi- neers take this approach, but their re- sults are not always perfect, which re- flects deficiencies in their understand- ing of the process. Quantitative measures of interac- tion (specifically the relative gain). Especially when models provide the basis for the process design, obtain- ing quantitative values for the degree of interaction is relatively easy. This
technology cannot be applied blindly, however. The trick is to understand what these numbers are telling us about the process and our ability to control it. It is not necessary to eliminate all interaction; it only is necessary to re- duce the degree of interaction to the point where the loops will deliver satisfactory performance. Introduc- ing ratios, sums, function generators, etc., also can prove beneficial in this endeavor. Model-predictive controllers have been suggested as the ultimate solu- tion to such problems. Unfortunately, their implementation requires consid- erable effort by specialists knowl- edgeable about the technology. Com- parable expertise also is required to keep them running. As originally pro- posed, model-predictive control was an enabling technology that permitted the benefits of plant optimization ef- forts to be realized. There certainly are applications where model-predic- tive control is the preferred (and pos- sibly only) solution, but always iden- tify the benefits (in economic terms) before proposing a model-predictive controller.
- Check for departures from lin- earity. A change in the manipulated variable will prompt a change in the controlled variable, the ratio being the sensitivity of the process. Will a given change in the manipulated vari- able lead to a small change under cer- tain conditions, but to a large change under other conditions? This will lead to tuning problems in PID loops, and to a degradation in the performance of a model-predictive controller. A change of 2:1 in the process sensitivi- ty will degrade the performance of the controls, but usually not to a suf- ficient degree to justify expending ef- fort to address it. Control techniques are available to address this problem, but each has its limitations: Cascade control. You must identify a dependent variable that has a linear effect on the controlled variable. Such a dependent variable may not exist.
P ROCESS C ONTROL
22 August 2000 Chemical Engineering Progress
Adaptive control. An understand- ing of the process is necessary to de- velop the relationship upon which the adaptive controller is based. Sched- uled tuning is a simplistic form of adaptive control, but it also requires some process understanding to be formulated properly. Self-tuning controllers. These po- tentially can respond to slow drifts in process sensitivities, due to equip- ment wear, fouling of heat-transfer surfaces, etc. They, however, cannot respond to rapid changes in process sensitivities. The best approach is to find a ma- nipulated variable to which the con- trolled variable is linearly related.
- Assess the impact of distur- bances. Where a disturbance has a major influence on a controlled variable, the performance of the control loop will suffer. There are two options: 1. Provide control logic that rapidly responds to the disturbance. The following are possibilities: Cascade control. You must identi- fy a dependent variable that is affect- ed by the disturbance more rapidly than the controlled variable is. Such a dependent variable may not exist. Feedforward control. The distur- bance must be measured and logic in- corporated into the controls to take corrective action before the distur- bance affects the controlled variable. Measuring the disturbance frequently is the major obstacle. 2. Eliminate the disturbance at its source. This often is the preferred ap- proach. When the raw materials are natural products, however, some vari- ability is inevitable.
- Consider constraints on the de- pendent variables. There is one situa- tion where the value of a dependent variable becomes of concern, namely, when it exceeds limits known as con- straints. The process controls often include logic specifically designed to respond should the value of the de- pendent variable approach the con- straint. For example, consider a pres- sure reactor. Provided the pressure is
below its constraint, the temperature in the reactor would be controlled. But, should the pressure reach the constraint, the controls switch from controlling the temperature to con- trolling the pressure. Such logic is re- ferred to as override control. If violation of a constraint is a haz- ard to people or equipment, it is the responsibility of the safety system to take whatever action is appropriate should the constraint be violated. For the pressure reactor, the pressure con- straint specified to the process con- trols is less than the setting on the pressure relief device. It is the pres- sure relief device, not the process controls, however, that ultimately is responsible for protecting people and equipment.
Control problems
in operating plants
If the P&ID is appropriate, the plant startup team should be able to successfully commission the controls, which includes such activities as con- troller tuning. When difficulties are encountered, there are several possi- bilities. One is that the P&ID is not correct — that is, the control strategy does not properly reflect the charac- teristics of the process. Basically, there are two alternatives for address- ing this:
1. The control engineers can be- come familiar with the process. Even if such individuals are on staff and currently are available, becoming fa- miliar with a complex process will take some time. If an outside expert is retained, it is usually necessary to re- veal proprietary technology. 2. The process engineers can be- come familiar with the principles of process control. This is where the “just-in-time” education that is possi- ble with computer-based technologies has distinct advantages. (Note the use of the term “education.” Training is learning to execute a predetermined sequence of steps. You are trained to change the tire on a car. Education in- volves mastering principles that can be applied to a variety of situations,
some of which are not even contem- plated at the time. You are educated to solve process control problems.) As processes become more complex, the latter alternative be- comes more attractive. Indeed, today, process engineers generally are in the best position to address these problems. The customary initial report of a process control problem is a phone or e-mail message that, in effect, states that some part (or perhaps all) of the controls do not work. If the initial re- port gives any useful information at all, consider yourself to be fortunate. As with any type of problem, the first step in its solution is to accurate- ly define the problem. It may be with the routine performance of the con- trols or with how the controls reacted to a specific event. In either case, we need to address two separate, but re- lated, issues:
1. What did the controls actually do? 2. What was expected of the controls? Presumably “it doesn’t work” means that there is a difference be- tween these two. A good starting point is to un- derstand what the operations per- sonnel expected the control system to do. This brings up a variety of possibilities:
- The controls responded differ- ently than they were designed to do. There is something in the implemen- tation of the controls or associated hardware (measurement device and final control element) that must be corrected.
- The controls responded as they were designed to do, but this re- sponse was not appropriate. The P&ID needs a thorough analysis.
- The controls responded in an appropriate manner. Operations per- sonnel need a better understanding of what control actions are appropriate. In this article, we’ll focus on the first two possibilities. Determining what the controls ac- tually did involves two aspects:
Chemical Engineering Progress August 2000 23
the valve have a larger effect on con- trol system performance when the valve is oversized. Occasionally, the converse is en- countered. As the capacity of the pro- cess is increased, flows also must in- crease. This can lead to some control valves being operated with the by- pass partially open at all times. Changing the opening of the bypass valve will affect the performance of the controls. As the mechanical components wear, actuators exhibit hysteresis. The packing required to prevent leaks around the valve stem resists move- ment of the valve stem, leading to a behavior called stiction. Both of these usually result in a cycle in both the controller output and the measured variable. When a flow measurement is available, the presence of these phenomena can be easily detected. Are there times when the controller is changing its output but the flow is not changing? If so, then the actuator is not positioning the valve to the value specified by the controls. Controller hardware. In the days of pneumatic and electronic analog controllers, hardware malfunctions within the controller itself were po- tential problems that had to be con- sidered. But, today’s digital hardware generally is either working perfectly or not working at all. With the self-di- agnostic features normally incorpo- rated into computer-based industrial control products, problems with the hardware usually are very obvious. Both pneumatic and electronic analog equipment are susceptible to significant errors in the calibration of one or more of the controller tuning adjustments. With digital controls, tuning-adjustment calibration errors are simply not possible. Process. The basic question is “Can the process actually do what the controls are expected to make it do?” With the “our system can do anything you want” sales hype from manufac- turers, people sometimes have unreal- istic expectations for the process con- trols. The process ultimately deter-
mines what is possible. And, it occa- sionally does this in some rather sub- tle ways that only can be appreciated from a thorough understanding of the process itself. Process engineers are clearly in the best position to sort this out. Controls not matched to process. The performance of the controls is determined by two factors:
1. the control structure, as repre- sented by the P&ID; and 2. tuning of the controls to the process. The tuning issue must be assessed first. Tuning. Tuning is the procedure by which the characteristics of the con- troller are adjusted so that the con- troller is “in tune” with the process. Tuning techniques have been avail- able since the 1940s. When digital computers were first applied to pro- cess control in the 1960s, automated tuning was one of the promises. Yet, despite considerable work and the availability of a number of commer- cial products, most loops still are tuned with the traditional trial-and- error or “knob twiddling” approach.
The developer of a P&ID does not need to know how to tune a PID con- troller. But, when analyzing a prob- lem loop in a plant, don’t assume that a loop is well-tuned, even if the in- strument technicians contend that it is. The analysis of control problems in an operating plant must involve an assessment of how well the controller is tuned. The first step is to determine the tuning objective for the loop. Figure 2 illustrates two possibilities for the response to a change in the set point: one being the quarter decay ratio (that is, the ratio of the first peak overshoot to the second is 4:1), and the other being critically damped (little or no overshoot). The tuning coefficients are given in Table 1. Instrument technicians are taught that good performance is a response with a quarter decay ratio. That the quarter-decay-ratio criterion does provide superior performance is illus- trated by the responses to a distur- bance or load change (tuning coeffi- cients are the same as in the previous example) shown in Figure 3. A good indication of loop perfor-
Chemical Engineering Progress August 2000 25
-15.0 0.0 15.0 30.0 45.0 60.
155
150
165
160
170
145
Temperature,
˚F Set Point
Critically Damped
Decay Ratio = 1/ 1.2˚F
4.8˚F
Figure 2. Two tuning approaches.
Table 1. Tuning coefficients for Figure 1.
Controller Gain, KC Reset Time, TI %/% min Quarter decay ratio 2.0 5. Critically damped 0.5 5.
mance is the maximum deviation from the set point (the magnitude of the control error at the peak). For our example, the quarter decay ratio gives a deviation of 4.4°F vs. 7.2°F for crit- ically damped. Production personnel, however, often are not comfortable with the de- gree of oscillations in responses with a quarter decay ratio. When tuned this way, increasing the sensitivity of any element in the loop by approximately a factor of two results in an unstable loop. For most production personnel, this is a little too close for comfort. When you are involved in plant oper- ations, you do not want your phone to ring at 2:00 in the morning because some loop is unstable. You want de- pendable controls, and you are will- ing to sacrifice performance to en- hance dependability. After all, you, not the instrument technicians enam- ored with quarter decay, will get that 2:00 am call. A widespread malady in con- troller tuning is for the reset time to be too short, because of the common misconception that fast response is attained through a short reset time. This is not the case at all. The con- tributions of the three modes are as follows: Proportional — determines speed of response of the loop. Integral or reset — forces the loop to line out at its set point (eliminates offset or droop). Derivative — enhances stability margin, which, in turn, permits the controller gain to be increased to ob- tain a faster response. Compared to the controller gain, the reset time has little effect on the speed of response of a loop. To obtain fast response, make the controller gain as high as possible, not the reset time as short as possible. A major deficiency with using the quarter-decay-ratio performance ob- jective is that, for PI and PID con- trollers, many combinations of the tuning coefficients will give a re- sponse with a quarter decay ratio. In the previous example, a controller
gain of 0.5%/% and a reset time of 5.0 min gave a critically damped re- sponse. Shortening the reset time to 1.8 min, however, gives a response with a quarter decay ratio. Figure 4 il- lustrates two responses — both have a quarter decay ratio, but the respons- es certainly are not equivalent. The fastest response is the one
with the higher controller gain, not with the shorter reset time. The bene- fits of the faster response are more clearly illustrated by the load re- sponses, as seen in Figure 5 and sum- marized in Table 2. Note that the faster response (and the one whose maximum deviation from the set point is the lesser of the two) has the
P ROCESS C ONTROL
26 August 2000 Chemical Engineering Progress
Time, min
-15.0 0.0 15.0 30.0 45.0 60.
155
150
165
160
170
145
Temperature,
˚F Set Point
KC = 2.0 %/% T (^) I = 5.0 min
KC = 0.5 %/% T (^) I = 1.8 min
Figure 4. Many combinations of coefficients give quarter decay ratio.
Time, min
-15.0 0.0 15.0 30.0 45.0 60.
155
150
145
140
Temperature,
˚F Decay Ratio = 1/
Critically Damped
Figure 3. Quarter decay ratio provides better response.
Time, min
-15.0 0.0 15.0 30.0 45.0 60.
155
150
145
140
Temperature,
˚F
A B
Figure 5. Faster response offers benefits.
loop is specified on a P&ID does not mean that, in fact, it will work. The control structure indicated on the P&ID might not properly reflect the process requirements. The symptom of this is one or more loops that can- not be satisfactorily tuned. The fol- lowing are the common process problems that lead to untunable loops:
- process nonlinearities;
- excessive dead time;
- improperly nested cascade; and
- interaction between controlled variables. It is easy to throw technology at such problems, resorting to automat- ic tuning, expert systems, model-pre- dictive controllers, etc. But, until the problem is understood and the P&ID revised, the loop will not perform satisfactorily. Process nonlinearities. A nonlin- ear process exhibits different behav- ior under different process conditions. Most processes are nonlinear with re- gard to process throughput. Process- es, however, can exhibit nonlinear be- havior with regard to other variables as well. Batch processes are particu- larly notorious in this respect. In such cases, the controller can be satisfactorily tuned at a point in time. But when process conditions shift, the tuning is no longer satisfactory. The logical starting point is to tune the controller with the process operat- ing at or near normal (or design) con- ditions. All processes, though, are subjected to major upsets from time to time. How does the controller per- form under the process conditions during these upsets? We normally learn the answer through operational experience. This is one source of the “it doesn’t work” messages. The first step is to determine the process con- ditions that led to the problem. The controller can be retuned to these conditions (the “one tuning fits all” approach). This, however, usually leads to some sacrifice in perfor- mance at the normal operating condi- tions. The alternative is to develop a more sophisticated control structure
that can cope with the changes in pro- cess characteristics. Dead time. The dynamic charac- teristic that presents the most difficul- ty for the PID control equation is dead time. Most processes exhibit dead time, but to a modest extent. This results in some deterioration in the performance of a PID controller, but not enough to justify implement- ing an alternative approach. As the dead time rises, the first mode to suffer is derivative. The ben- efits from derivative decrease as the dead time increases. The next mode to suffer is proportional. As the dead time goes up, the controller gain must be reduced to maintain an acceptable margin of stability. For long dead times, the PID controller essentially will reduce to a pure reset controller. The controller may operate on auto- matic, but the response is so slow that the controller is practically useless. For processes with a long dead time, a model-based control technolo- gy known as dead-time compensation is available and has been routinely ap- plied to paper machines since the late 1960s. Model-predictive controllers also provide dead-time compensation. There, however, is one prerequisite to the successful application of dead- time compensation: a good value must be available for the process dead time. In paper machines, the dead- time compensator is, in turn, compen- sated for the effect of changes in ma- chine speed on the dead time. Improperly nested cascade. In a cascade control configuration, one controller (the outer controller) pro- vides the set point to another con- troller (the inner one). For this con- figuration to perform satisfactorily, the dynamics of the inner loop must be faster than the dynamics of the outer loop. The usual desire is that the inner loop be at least five times faster than the outer loop. An inade- quate separation of dynamics leads to tuning problems in the outer loop. When the cascade is a temperature controller providing the set point to a flow controller, such a separation in
dynamics almost always is assured. But, when a temperature controller provides the set point to another tem- perature controller, the separation of dynamics may not be adequate. Fur- thermore, temperature processes al- ways are interacting stages (each temperature affects the other temper- ature), which complicates the inter- pretation of the responses in the inner loop temperature. These control con- figurations have to be analyzed by someone very familiar with the char- acteristics of the process. Interaction. All processes are multi- variable in nature. If ten variables are to be controlled, most P&IDs will pro- pose to do this with ten individual loop controllers that are paired with ten final control elements. A single-loop con- troller works well in a multivariable environment, however, only when the following statements are true:
- The final control element for the loop does not affect the measured variables for the other loops.
- None of the final control ele- ments for the other loops affect the measured variable for this loop. When these statements are not true, the result is loop interaction. A small degree of interaction can be tol- erated, but even modest degrees of interaction will degrade the perfor- mance of one or more loops. Some- times, the interaction can be suffi- ciently reduced by incorporating ra- tios, summers, function generators, or other simple elements into the control configuration. For the more severe situations, a model-predictive con- troller may be the best (or perhaps only) solution.
P ROCESS C ONTROL
28 August 2000 Chemical Engineering Progress
Upcoming AIChE control courses
In early November in Los Angeles, Dr. Smith is scheduled to give courses on “Auto- matic Control of Processes,” “Control of Batch Processes,” and “Distillation Control.” For more information on these courses, or to re- ceive a catalog on all AIChE courses, contact Nina Weber, Director of Education Services ((212) 591–7526; E-mail: [email protected]).
The role
for process engineers
Process control issues can be di- vided into two categories: Systems issues. These include con- figuring the hardware, building points (for inputs, outputs, control points, etc.), creating graphic displays for the process operator, and so on. This “grunt work” can be done with very little, if any, understanding of the pro- cess. The systems issues are ade- quately covered by training courses (yes, training, not education) offered by suppliers. Process engineers cer- tainly can do this if they are so in- clined, but these tasks should be left to others to free the process engi- neers’ time for the more important as- pects of the control system. Process issues. At the design stage, these center on creating the P&IDs for the process controls. Dur- ing and after startup, these involve re- solving problems with the process controls. Process issues are the im- portant aspects of the control system, and it is crucial that they be done cor- rectly. Both endeavors require a sub- stantial understanding of the process — so, process engineers are in the best position to do them. Yet, a lot of process engineers do not feel comfortable about addressing the process issues. Many of these en- gineers probably owe their misgiv- ings to the “process control” course they endured at university — it more than likely emphasized mathematics and gave the wrong impression about the practice of process control. To address process issues requires an appreciation of the following concepts: Process characteristics. The focus should be on how steady-state sensi- tivities and simple process dynamics (integrating vs. nonintegrating, time constants vs. dead times, etc.) impact control loop performance. PID control equation. Topics in- clude the role of the three modes, the options available in digital systems (especially tracking features), and tuning. Actually, P&IDs can be de- veloped with very little understanding
of tuning; tuning must be appreciated, however, by those who resolve con- trol problems in the plant. Input/output (I/O) elements. This is limited to those aspects of mea- surement devices and final control el- ements that affect control system per- formance. Measurement device selec- tion (thermocouples vs. RTDs vs. py- rometers vs. …) and valve sizing are not required. Some familiarity with filtering and smoothing is needed so that the engineer will not hesitate to remove them. Split range. This handles situations where, under some conditions, a vari- able is controlled using one manipu- lated variable, but, under other condi- tions, using a different manipulated variable. Most plants have a couple of such applications. Override control. This technology is employed to address the constraints on production operations. The track- ing issues must be understood. Cascade control. The advent of digital controls has led to an increase in the application of cascade control. Process engineers are in the best posi- tion to recognize opportunities for this technology, and, probably, will have to tune most of the temperature- to-temperature cascades. Ratio control. This simple technol- ogy can be understood quite quickly. Feedforward control. Most process engineers need only an introduction to this topic, although a few should have an in-depth exposure. Multivariable control. All process engineers must understand how inter- action can lead to problems in control systems, and how to recognize when interaction is the source of controller tuning difficulties. Some process en- gineers must know how to quantita- tively assess the degree of interaction and how to develop control configu- rations that have acceptably low de- grees of interaction. Topics such as model-predictive control can be left to the control experts, though. Let’s conclude by putting the proper perspective on mathematics. The PID control equation is a differ-
ential equation; the dynamic charac- teristics of processes are described by differential equations. Engineers do not need to solve either. The control systems solve the PID control equa- tions; various simulation packages are available to solve the differential equations that describe the process. But, in process control, it is necessary to work with differential equations. This is where the Laplace transform enters; as noted previously, differen- tial equations are very conveniently represented in the Laplace domain. It is understandable why a process engineer’s initial reaction would be to attempt to avoid the Laplace trans- form entirely. The above topics can be presented without using the Laplace transform at all. Laplace ex- pressions, however, actually make the presentation both easier and clearer. In the final analysis, it is better to join them than to fight them. But, joining them does not require becoming an expert in the subject. Many of the topics (such as partial fraction expansion) taught in college courses never are used in the field. What a process engineer needs to know about Laplace transforms can be conveyed in 15 min. Instead of trying to avoid the subject entirely, it is best to invest the 15 min. Then, you will not be intimidated by the Laplace expressions that occasionally appear in the manuals provided by control system manufacturers. And, it only takes a little knowledge of Laplace transforms to keep the con- trol engineers honest. CEP
Chemical Engineering Progress August 2000 29
C. L. Smith is President of Cecil L. Smith, Inc., Baton Rouge, LA ((225) 761-4392; Fax: (225) 761–4393; E-mail: [email protected]), a firm specializing in automation technology. He is the author or co-author of 14 books and over 100 articles. He received BS, MS, and PhD degrees in chemical engineering from Louisiana State University.
