Anda di halaman 1dari 4

instrumentation

When is PID
Not the Answer?
Social media group member posits the question: How come PID is the
answer to so many applications? Is it because advanced control is rarely
needed or that control engineers are not pushing the envelope?

P
David Greenfield ID has long been one of the more is Advanced Control good for?” a high level of
Control Engineering popular topics covered at Control response was assured.
Engineering, and for good reason— Kodati noted that, while a graduate stu-
our subscribers see it as an inte- dent, he was “amazed by the subject of con-
gral component of their work. The trol and couldn’t wait to see the applications
popularity of the topic among control engineers of the theory applied in industry.” Three years
worldwide does not appear to be waning either. after graduation, however, he began to wonder
Looking back over the past two years, loop tun- why PID seemed to be the answer for so many
ing and PID-related articles are consistently applications.
among the most popular articles visited on the Members of the social media group quickly
Control Engineering Website. responded, and continued to do so for months
So when Parasar Kodati, application support after the initial post. Their responses largely
engineer at the MathWorks, posted this question broke out along three lines: why PID is right for
on the Automation & Control Engineering Linke- so many applications, when advanced control is
dIn site: “When PID is not the answer: What more suitable, and how to analyze a situation to
determine if PID or advanced con-
Standard PID algorithm trol is the best option.

When PID is enough


Disturbances “From my experience, a great many
1 systems have very simple dynam-
Setpoint Error ics, and in these situations PID is
1 + often sufficient to provide the per-
+
--
Kp
Ti ∫ +
+
+
Process +
formance needed,” said Doug Harri-
d
man, senior mechatronics engineer
Td dt at Hewlett-Packard. “PID control-
lers can be more intuitive to tune.
Controller It’s easier to reason out the expected
behavior when changing one of the
PID gains than it is to think about
Process variable pole placement.”
He also noted that, since most
Source: Control Engineering
engineers “never take any advanced
control theory,” this could be anoth-
The standard or ISA standard form of the PID algorithm allows a user to simultaneously er reason behind PID’s popularity.
adjust the contributions of all three terms by means of a single tuning parameter (Kp). The Kareem Suhwail, systems engi-
proportional, integral, and derivative actions all depend on the value of Kp; hence the alter- neer at RoviSys, says that, from
native label dependent or dependent ideal PID. This arrangement is occasionally labeled the an automation perspective, PID is
academic algorithm even though academicians rarely use it. Source: Control Engineering, more than enough for 99% of con-
“Sorting Out PID Controller Differences” by Vance VanDoren, February 2009. trol situations. “Realize that, while
54 ● MARCH 2010 CONTROL ENGINEERING ● www.controleng.com
Automation & Control
studying advanced controls, you are often given social media forums
transfer functions, block diagrams, state equa- To become a part of the conversation or
tions, etc., it is virtually impossible to model merely follow the “Automation & Control”
a running production line with these and it is discussions in real-time, access the forums on
unrealistic to perform an expensive advanced Facebook and LinkedIn via the following links:
control analysis for the small gains it would give • Automation & Control Engineering group on
you over PID,” he said. LinkedIn: http://tinyurl.com/y8tckah
“Most control loops in the chemical indus- • Automation & Control group on Facebook:
try, while non-linear to some degree, are linear http://tinyurl.com/yk43xmf
enough in the normal control region near set-
point” for which PID works fine, said Joseph
Alford, a consultant with Automation Consult- [advanced] control strategies and translate it into
ing Services. It may not be not optimal control, a real return on investment.”
but it’s good enough.
Alford also echoed Harriman’s comments Alternate options
that most engineers typically do not take more While most respondents agree that PID is good
than one undergraduate control course and so enough for most control applications, what about
“have only been trained on PID and a few relat- those instances when it’s not good enough?
ed simple control configurations (e.g., cascade).” What then are the best options engineers should
Add to the education factor PID’s simplicity and learn more about in order to consider them for
off-the-shelf availability, and you have three application?
very big reasons for PID’s dominance in control “PID only, in its simplest form, is not enough
applications. for position control for extremely dynamic or
In another part of his comment, Alford second order systems,” said Richard Meyer-
referred to a survey he had seen which indicated hoefer, sales manager at Delta Computer Sys-
that audits of many industrial control loops using tems Inc. “Advanced algorithms are available
model-based or other complex control strategies that will create a system model and automati-
often had the loops set in manual mode. Alford cally calculate additional gains such as velocity
said this was often the case after the engineer feed forwards, acceleration and deceleration feed
who designed the advanced control strategy had forwards, jerk feed forwards, and double deriv-
moved on to another assignment. This indicates atives. These advanced gains are based upon
that, for most engineers, it is “difficult to main- natural frequencies. PID is used to correct for
tain continuous support of complex control loops non-linearity. Feed forwards should be able to
when new support people are not familiar with correct for 90% of the error. In these cases, [only
the control algorithm. There are probably also using] PID leads to very unstable systems.”
validation issues regarding complex control algo- Meyerhoefer adds that it is often a lack of
rithms in highly regulated industries such as understanding on the part of many engineers
pharmaceutical, biotech, medical devices, and that lead them to abandon closed loop control
nuclear,” said Alford. options in favor of more mechanical options.
Though many industrial processes are Unfortunately, in many instances, these options
becoming more challenging, Thomas Swift, an often lead to a lack of repeatability, lower resolu-
industrial automation and control profession- tion, and increased down time.
al, says that PID remains his tool of choice for PID coupled with Smith Predictor can address
most of the control issues he encounters. Even simple transport delays easily, according to Ser-
in many advanced applications, “classical meth- gei Kuznetsov, principal control systems engineer
ods such as feed forward and cascaded feed- at The Automation Group. He noted that he has
back are, in my experience, much preferred to seen Smith Predictor used for a wood chip level
increasing the complexity of the core control control that had been working in manual mode
algorithm,” he said. for more than 20 years. “The client ended up
The secret to the success of PID-based con- saving $1.5 million dollars in raw material costs
trol architectures (such as ratio control, cascade annually (about 7%),” Kuznetsov said. ONLINE
control, and feedforward control), according to “Model-based control is clearly a superior To access all Control
Massimiliano Veronesi, product marketing man- approach, from the standpoint of responsive- Engineering PID-related
ager at Yokogawa, are the low cost/high benefit ness, following errors and settle time, in those content, input “PID” into the
ratio they provide. “It is not always easy to mea- situations where it can be applied,” said Michael search box atop any page at
sure the performance improvement provided by Thompson, principal development analyst at The www.controleng.com.

www.controleng.com ● CONTROL ENGINEERING MARCH 2010 ● 55


instrumentation

There are also many situations where non-lin-


The ideal form of the PID formula
ear gain scheduling can be effectively employed
in a PID control-loop without additional sensors
or significant complexity, according to Thomp-
son. In such cases, the gain-scheduling may sim-
ply be a transfer function that maps a non-linear
variable to a more linear one. In these cases, the
PID controller would act the same.

CO(t) is the controller’s current output; e(t) = SP - PV (t) is the Application advice
error between the set point (SP) and the process variable Regardless of whether PID is adequate for most
PV(t); P is the controller gain; T1 is the integral time TD is the applications or not, the only thing that real-
derivative time. Source: Control Engineering, “Loop Tuning ly matters from an engineering perspective is
Fundamentals,” by Vance VanDoren, July 2003. which control method will provide the optimal
result for the particular application the engineer
Timken Co. “However, model-based control often happens to be working on at the time. This next
requires additional terms in the feedback loop. set of advice from practicing engineers addressed
For example, if I am trying to change the temper- just that.
ature of a volume of liquid in a storage contain- As a starting point, Igor Begelfor, instrument
er, it definitely helps to know how much liquid package engineer at AMEC, suggests that when
is present in addition to the temperature of the PID does not work, start looking at the field mea-
liquid. Charging a battery is a similar problem. surement and control devices, as they have likely
It helps to know how much charge it already been misapplied. Once you’ve eliminated these
has. Sometimes that information is not so easy to as the problem and concluded that PID may not
determine from the available sensors.” provide the optimal control for your project, it’s
time to start looking closely at the application.
Recalling work on systems with extensive
Don’t forget advanced control basics higher order dynamics that needed to meet
stringent stability and tracking requirements,
While it’s tempting to apply any If a loop has large dead time, then Hai Ho, vice president of engineering and prod-
number of advanced control meth- you won’t get great control from uct development at HID Global, noted his use of
ods to optimize a given process any feedback controller because multi-variable state space, adaptive, and learning
to its ultimate in performance, the feedback control is like driving control methods rather than PID. The systems
reality is that most processes—and your car by looking in the rearview requiring this advanced level of control were: a
process owners—do not require mirror. digital flight control system for satellite missiles;
that level of control. If the control • Most plant personnel don’t and controlling the position of a magnetic read/
method can’t deliver in terms of know how well their existing con- write head to the micro-inch accuracy under
cost and quality improvements, it’s trols are performing. “I’ve person- wind disturbance conditions equivalent to a cat-
usually not the answer you need. ally been in hundreds of plants, egory 4 hurricane.
To keep your control method and typically 30% of PID loops Offering a more practical example using non-
decisions as useful as possible, are running in manual mode, and linear gain scheduling, consider the problem of a
George Buckbee, vice president nobody knew it,” said Buckbee. He scanning CMM (coordinate measuring machine),
of marketing and product devel- adds that control loop monitoring suggested Michael Thompson, principal develop-
opment at ExperTune Inc., offers software can be used to diagnose ment analyst at The Timken Co. “During scan-
these four checkpoints: and give visibility to instrument, ning of the surfaces, the axes of the machine are
• Realize that the control hard- valve, and control performance. controlled using feedback from a probe. How-
ware in an existing plant is often • Focus on the root cause of ever, most probes only provide feedback over a
in bad condition. A typical process process upsets. “I have seen many very narrow range. A contact-based probe cannot
plant can expect about 30% of plants installing model predictive provide useful feedback when it is not in con-
control valves to have mechanical control on distillation columns, tact with the surface or when it is over deflected.
issues, says Buckbee. “If you have when the fundamental problem is Also, these machines generally try to scan the
sticky control valves, noisy instru- process upsets coming from utili- surface with minimal probing forces (for reduced
mentation, or poor installations, ties. You’ll often get a better return wear, less deformation of the surface, and great-
then it doesn’t really matter what on investment from stabilizing the er accuracy). So the control problem is highly
type of controller you have. You boiler plant operations,” he said. asymmetric,” he said.
will still have poor performance.” The trick is to be able to analyze “Near the target force/deflection (when error
• Have reasonable expectations. the data for root cause. is nearly zero), acceptable and stable control can
be achieved with linear gains and standard PID
56 ● MARCH 2010 CONTROL ENGINEERING ● www.controleng.com
Tank Gauging Problems?

control. But when the probe is undeflected, the error signal


is low (because of the low target force) so you must rely on
heavy integral gain in order to achieve a reasonable approach LIQUID LEVEL COMPUTERS
velocity,” Thompson said. “That heavy integral control is not
necessarily desirable once the probe is in contact with the
part. In fact, the integrator wind up in such a case can lead to The LevelCom 100 uses pulsed bubbler
over-deflection of the probe during overshoot after the probe depth sensing technology, which resolves
makes contact. Likewise, if the probe approaches an over- many of the problems long associated
deflection condition, it is desirable to drive it much harder with continuous flow bubblers and
than normal in order to avoid the over-deflection.”
reduces air consumption.
One technique that Thompson said he has used to address
these problems is to “apply control action that is proportional
to the cube of the error term in addition to the normal (linear)
proportional control. The cube is used to maintain the sign of
the error term. To address the asymmetry, I chose to toggle
the gain sets depending on whether the probe was minimally
deflected or not, but a quadratic term could also be used to
create asymmetry. One reason why I have preferred to toggle
the gain sets is due to the fact that the damping of the probe
changes when it is in contact with the surface.”
Ultimately, Thompson says the gains are chosen so that
there is a linear region around zero error. “Via non-lineari-
ty (cubic and quadratic for example), the transfer function is
made much stronger than normal when the probe approach-
es an undeflected or over-deflected condition,” he said. “The
stronger response in these regions would not be stable if it
was applied linearly throughout the range of probe deflection.
Applications can be in a vented or pressurized
Therefore, the zero-error condition becomes a region of stabil-
ity where gains are sufficiently low to avoid instability.” tank the liquid must be able to tolerate air or
nitrogen injection. Typical applications include oils,
Final analysis fuel, potable water, viscous fluids, slurries, acids,
Helping to encapsulate this wide-ranging discussion, one post wastewater and foaming liquids.
in particular got to the core of the issue as it pertains to the
day-to-day duties of a control engineer.
Thomas Stevic, a controls engineer at DMAX, said that Features include:
when designing a solution to a problem, he finds that he typi-
cally has two masters to serve. The first master is the person • Field configurable with multiple set points
who is responsible for the output of his design. This is often and specific gravity corrections
“the person who pays the bills,” Stevic said. “These people
usually do not care how a machine or process works. Their • Integrates with third party controllers, DCS,
concern is: Does the machine make what it is supposed to data acquisition systems, and software
make? While exotic solutions may increase quality, decrease
production time, make better use of resources and, from an • Modbus connectivity standard with DeviceNet
engineering viewpoint, be very cool to use and inch closer to and Profibus available
perfection, their use is not often a benefit to the first master.”
The second master is described by Stevic as the person or • Diagnostic capabilities detect leaks, plugged lines,
people who will “maintain the system after I am gone. Ulti- pressure problems, and hardware errors.
mately, my designs have to address the least capable people
who may be required to maintain or modify my design.”
Given the requirements of the second master, Stevic says
that, given a choice between using adaptive neural network TMS
control or a thermostat, if the thermostat will work, that is 6040 N. Cutter Circle, Suite 302
what he will choose. “If neural nets or fuzzy logic are required Portland, OR 97217
to produce the desired result, I will use them, but only after P: 503.285.8947
examining the simpler solutions,” he said. ce F: 503.285.1379
David Greenfield is editorial director. Reach him at david. www.levelcom.net
greenfield@reedbusiness.com.
www.controleng.com ● CONTROL ENGINEERING MARCH 2010 ● 57

Anda mungkin juga menyukai