Professional Documents
Culture Documents
Control Engineering
for Chemical Engineers
Chemical engineers who are aware of process control requirements and challenges are in a
position to improve process designs
C
Lou Heavner hemical engineers SODT vs INPUT
Emerson are ideal candidates (Non-integrating)
1.0
for control engineer-
ing jobs. They under- 0.9
IN BRIEF stand processes and process 0.8
Percent change in process output
Design basis and variability frequency (“fast”) variability (less than a FIGURE 2. Typical first-order plus deadtime
Process engineering focuses on process few seconds). (FODT) responses are characterized by a
rapid initial response to a process input,
design, and defines or assumes a de- Fortunately, process design can often followed by slowing response as a new
sign basis. That basis typically includes be used to attenuate fast variability. steady state is reached
normal, maximum and minimum pro- Surge vessels can be used to attenuate
duction rates, and the process engineer highly variable flows between units, for
tries to optimize the process design, first example, reducing the disruption to the
in terms of capital cost and second in downstream unit from variability in the
terms of operability. At this stage, project upstream unit.
cost considerations and the availability Control engineers need to understand
of standard process equipment may re- process dynamics, a topic area that is
quire design compromises that lead to a not always considered as part of the
process design with control challenges. core of process design. It is convenient
The design basis is a guideline, but to think of process dynamics in terms
operating conditions in a commissioned of process inputs and process observa-
plant may change over time. Equip- tions. Process inputs are material or en-
ment (especially control valves) wear, ergy flows, and they may be flows into,
feedstock qualities vary, catalysts age, out of, or intermediate within a given
processes are impacted by varying am- process. As flows are changed, the
bient conditions, and other sources of process is affected, as seen by process
variability impact production. Market observations. Process observations are
and regulatory conditions may also measured as variables like tempera-
vary, shifting demand for certain prod- tures, pressures, levels, compositions
ucts and byproducts or penalizing the and flowrates.
production of waste products. The con- As process designs are optimized
trol system of the plant is intended to for energy recovery and minimization
mitigate the effect of incoming sources of both capital cost and operating cost
of variability on product quality variabil- for a plant, they incorporate increasing
ity. As plants become increasingly com- integration between process streams.
plex, operators are faced with bigger If variability is not controlled in a highly
challenges, and simply operating the integrated process with a high degree
process manually is no longer an op- of process interactions, there are more
tion. A frequently cited analogy is the pathways for it to create quality issues.
pilot in an advanced meta-stable jet de- Therefore, it is increasingly important for
pendent on advanced avionics. design and control engineers to work
Perhaps the best way to look at au- together to ensure operability and strat-
tomation and control is as the business egize how to attenuate variability.
of managing process variability in real-
time. One important thing to understand Control basics
is that control systems are generally able Most process responses can be classi-
to attenuate low-frequency (“slow”) vari- fied into self-regulating and non-self-reg-
ability (on the order of seconds, minutes ulating (or integrating). Self-regulating pro-
or more), but cannot attenuate high- cesses respond to a change in process
P
ID controllers are defined by the control algorithm, which gener- action is sensitive to noise in the error, which magnifies the rate of
ates an output based on the difference between setpoint and change, even when the error isn’t really changing. For that reason,
process variable (PV). That difference is called the error, and the derivative action is rarely used on noisy processes and if it is needed,
most basic controller would be a proportional controller. The error is then filtering of the PV is recommended. Since a setpoint change
multiplied by a proportional gain and that result is the new output. The can look to the controller like an infinite rate of change and pro-
proportional gain may be an actual gain in terms of percent change of cesses usually change more slowly, many controllers have an op-
output per percent change of error or in terms of proportional band. tion to disable derivative action on setpoint changes and instead of
Proportional band is the same as gain divided by 100, so the effect is multiplying the rate of change of the error, the rate of change of the
the same, even if the units and value are different. When tuning a control PV is multiplied by the derivative term.
system, it is important to know whether the proportional tuning param- There are two steps to tuning a controller. First the process dy-
eter used in the controller being tuned is gain or proportional band. namics must be identified. This can be done with an open-loop or
When the error does not change, there is no change in output. This closed-loop step test. In open loop, the controller is put in manual
results in an offset for any load beyond the original load for which the mode and the output is stepped. The PV is observed and the pro-
controller was tuned. A home heating system might be set to control cess deadtime, gain, and time constants are estimated. Several
the temperature at 68˚F. During a cold night, the output when the steps should be made to identify any nonlinearity and to ensure the
error is zero might be 70%. But during a sunny afternoon that is not response is not being affected by an unmeasured disturbance. In
as cold, the output would still be 70% at zero error. But since not closed loop, the controller is forced to oscillate in a fixed cycle by
as much heating is required, the temperature would rise above 68˚F. stepping the output, forcing it to oscillate with an amplitude that
This results in a permanent off-set. will be dependent on the process gain and step size. This can be
Integral action overcomes the off-set by calculating the integral of achieved with a controller by zeroing the integral and derivative
error or persistence of the error. This action drives the controller error terms and adjusting the proportional gain until the cycle is repeat-
to zero by continuing to adjust the controller output after the propor- ing, or by using logic that switches the output when the cycling PV
tional action is complete. (In reality, these two actions are working crosses the setpoint value.
in tandem.) The integral of the error is multiplied by a gain that is The second step is calculating the tuning parameters. There are
actually in terms of time. Again, different controllers have defined the different guidelines proposed by different authors and even software
integral parameter in different ways. One is directly in time and the that will calculate the tuning parameters for the tuner to achieve
other is the inverse of time or repeats of the error per unit of time. the desired response. One guideline that is wisely favored is the
They are functionally equivalent, but when calculating tuning param- “lambda” tuning method. Lambda refers to the closed loop time
eters, the correct units must be used. Adding further complication, constant in a controller response. The advantage of this kind of tun-
the time can be expressed in different units, although seconds or ing is that the tuner is free to choose the speed of response or the
minutes are usually the design choice. aggressiveness of the controller tuning. There is a tradeoff in loop
And finally, there is a derivative term that considers the rate of tuning. As noted earlier, faster response or more aggressive tuning
change of the error. It provides a “kick” to a process where the error may result in some overshoot or even cycling response that is un-
is changing quickly and has a gain that is almost always in terms of desirable and the loop could become completely unstable if there is
time. However, again the units of time may be seconds or minutes. any nonlinearity in the process. Therefore, robustness is the sacrifice
Derivative is not often required, but can be helpful in processes that for more aggressive control and lambda can be used to strike an
can be modelled as multiple capacities or second order. Derivative optimal balance between robustness and aggressiveness.
input by settling into a new steady-state more steam enters the heat exchanger.
value. For example, if steam is increased First, the steam pressure in the exchanger
to a heat exchanger, the material being rises and heat transfers to the tubes and
heated will rise to a new temperature. Re- finally to the colder stream. The tempera-
ponses often take the approximate form ture of the cold stream takes some time
of first order plus deadtime (FODT) or before it begins to rise. Then it rises grad-
FIGURE 3. Integrating, or non-self-regu-
lating process variables do not settle into
second order plus deadtime (SODT) (Fig- ually and increases its rate of change until
a new steady-state value within allowable ures 1 and 2). In a heat exchanger, for ex- it approaches the new steady-state tem-
operating limits ample, when the steam valve is opened, perature, where the temperature rise be-
gins to slow. The characteristic response
SODT vs INPUT
(Integrating) is a SODT.
0.24 The steam flow began increasing as
0.22 soon as the valve started moving. But if
0.20 a controller was telling the valve to open,
Percent change in process output
tolerances.
Eliminate Downtime & Minimize Maintenance Costs
Reduce Energy Consumption & Emissions Visit our NEW website: http://hexoloy.com
For more information, technical data & case studies contact:
Maximize Productivity
SCD.sales@saint-gobain.com ■ 716-278-6233
Circle 34 on p. 90 or go to adlinks.chemengonline.com/66426-34
cals and specialty chemicals, rely heavily batch quality. There is a standard de-
on batch processing. Some engineers fined (ANSI/ISA S88) for batch process
muse that all processes are batch pro- control that standardizes the concepts
cesses, but some batches are longer of control, equipment and unit modules
than others. A batch control engineer in a batch process.
might suggest that a batch process is And batch processes are often de-
just a continuous process that never signed to make a variety of products or
gets the chance to reach steady-state. product grades. Furthermore, there may
Both are valid points of view. Designing be multiple trains of equipment with some
batch process sequences and recipes common process equipment or utilities.
fall right in the comfort zone for chemi- These plants may involve special recipes.
cal engineers. But the more interesting Recently in batch control, the focus has
part of batch control is not defining the been on managing multiple recipes and
normal sequence of steps. Rather, it is optimizing equipment selection for maxi-
defining what should happen when an mum or optimum production.
abnormal event occurs. Can a batch be Because product flaws in the phar-
“saved” following an upset or must it be maceutical industry can be devastating,
scrapped? What is required to rework a traceability is a major concern. This in-
batch that suffered an abnormal upset? cludes traceability of the materials con-
Thinking through the possible problems sumed in the production of pharmaceu-
that could disrupt a batch process and ticals, and traceability of the equipment
defining safe sequences to abort or re- and processes used to produce the
cover a batch are classic chemical engi- pharmaceuticals. Regulatory involve-
neering exercises. ment is high, and validation is an inte-
Another opportunity for batch optimi- gral part of pharmaceutical processes.
zation involves trying to minimize transi- This requires more data collection and
tion times between steps. This can be more rigorous adherence to manage-
done with equipment selection, but also ment of change (MOC) procedures than
with logic in the batch sequence. Ramp most other processes, whether batch
rates and dwell times can be minimized or continuous.
to the extent practical without impacting
48 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017
Process safety in control pact is data management and analysis.
Another area of process control deals Control systems have access to a great
with safety instrumented systems (SIS). deal of data other than control data. His-
Up to this point, the discussion has torization and archiving of process data
centered on control requirements to enables process engineers to identify
keep the process running in the face of and prioritize continuous improvement
variability. Safety systems have a single opportunities and allows management
function, which is to safely shut down to make more effective decisions regard-
a process if a catastrophe is imminent. ing operation and future investment.
Process engineers may be better pre- Business systems that manage main-
pared to consider process safety than tenance processes, quality processes,
most disciplines, at least with regard to planning processes and other work pro-
the CPI. The general concept is to evalu- cesses can be integrated with process
ate the risks, in terms of probability, and control. This has been enabled by modern
the magnitude of the consequences. technology for networking, databases,
Layers of protection are defined and operator interfaces and enterprise-man-
deployed to reduce the risk of a serious agement software all working together.
safety or environmental exposure. High- While the nature of integration of these
risk possibilities need to employ engineer- various systems requires more knowledge
ing solutions to reduce the risk. Some in computer programming, database ad-
solutions will include process design, ministration and networking than chemical
such as dikes around tanks and pressure- engineers might learn in their academic
relief equipment. Controls will also be em- programs, the process management re-
ployed to reduce the risk, including safety quirements require an understanding of
interlocks. The requirement for high on- the process plant and its economic sen-
demand availability of the safety-protec- sitivities. Chemical engineers are likely to
tion systems leads to specialized safety have a better understanding than most
systems with redundancy (including triple of the information required by company
redundancy) and pro-active diagnostics managers at both the local and corpo-
to monitor the health of the safety sys- rate level in order to make best use of the
tems. One of the first layers of protec- data and systems in place. The increasing
tion is alarm management, although it is wealth and richness of data makes analy-
limited by the presence of the human el- sis of that data with evolving “big data”
ement to respond to an alarm (for more tools a real opportunity. Networking, data
on alarm management, see Chem. Eng. sharing, and collaboration between the
March 2016, pp. 50–60). Designing plant and specialized resources located
safety-instrumented control systems is a far away is the promise of the Industrial
specialized area that is critical in managing Internet of Things (IIOT).
the risk of hazards in the process. There is
a growing trend to design safety systems Concluding remarks
to be integrated into — but still separate Often the greatest knowledge gap for
from — the basic control systems. Care a chemical engineer who wants to be-
is taken during design to ensure the inte- come a process automation engineer
gration does not create a vulnerability or is deep knowledge of instrument and
common point of failure of the safety sys- control hardware. This is not an insur-
tem function and reliability. mountable problem, however, because
Along with this trend is the increasing vendors are happy to share the infor-
use of diagnostics and capabilities of mation you need. A good salesperson,
“smart” instruments and field devices to perhaps contrary to popular opinion,
reduce the probability of failure on de- can be a valuable and trusted advisor.
mand. This is a critical consideration for The best salespeople know that exag-
safety systems, because they may not gerating the benefits of one offering for
be employed for long periods of time, immediate sale may win the order, but
if ever, but then must work when called will lose the confidence and trust of the
upon to shut a process down safely. customer for future opportunities.
In most cases, vendors truly do want
Process data to recommend the most economical so-
A final area of process control where pro- lution. To do that, they need to under-
cess engineers can have a significant im- stand the process and control require-
Circle 11 on p. 90 or go to adlinks.chemengonline.com/66426-11
CAS-236A.indd 1 1/24/17 1:18 PM
50 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017