Professional Documents
Culture Documents
TEMPERATURE
& PRESSURE
www.ControlGlobal.com
TABLE OF CONTENTS
Insights to process and loop performance 6
Full Page Ad
Bleed:
8.125”
10.75”
Powered by Possibility
For more information visit our website at www.watlow.com
www.ControlGlobal.com
AD INDEX
ACROMAG • www.acromag.com/sp 20
Krohne • www.us.krohne.com/optitemp-tt-51 10
Moore • www.miinet.com/THZ3TDZ3 15
Newark • www.newark.com 5
Watlow • www.watlow.com 3
H
ere we look at a myriad of metrics associated with a decrease in process effi-
on process and control loop per- ciency. Also an increase in the metrics for a
formance and show how to see particular part of a process may decrease
through the complexity and diversity to the metrics for other parts of the process.
recognize the commonality and underly- The following example cited in the April
ing principles. We will see how dozens of 2017 Control Talk Column “An ‘entitlement’
metrics simplify to two classes each for the approach to process control improvement”
process and the loop. We also provide a is indicative of the need to have metrics and
concise view of how to compute and use an understanding for the entire process:
these metrics and what affects them.
“In a recent application of MPC for thermal
Let’s start with process metrics because oxidizer temperature control that had a
while as automation engineers we are compound response complicating the PID
tuned into control metrics, our ultimate control scheme, there was a $700K per
goal is improvement in the process and year benefit clearly seen in reduced natu-
thus, process metrics. The improvement ral gas usage. However, the improvement
in profitability of a process comes down also reduced steam make to a turbo-gen-
to improving process efficiency and/or erator, reducing electricity generated by
capacity. Often these are interrelated in $300K per year. We reached a compromise
that an increase in process capacity is often of about $400K per year in net benefit
because of lost electrical power generation operations, the time period is usually the
from less steam to the turbo-generators. cycle time of a key phase in the batch and
We spent many hours to align the benefit may simply be the totals at the end of the
with measureable accounting for the natural phase or batch. For continuous operations,
gas reduction and the electrical purchases. I favor a time period that is an operator
Sometimes the loss of benefits is greater shift to recognize the key effect of oper-
than expected. You need to be upfront and ators on process performance. This time
make sure you don’t just shift costs to a dif- period is also suitable for evaluating other
ferent cost area.” sources of variability, such as the effect of
ambient conditions (day to night operation
Process efficiency can be increased by and weather) and feeds and recycle and
reducing energy use (e.g., electricity, steam, heat integration (upstream, downstream
coolant and other utilities) and raw mate- and parallel unit operations). The periods
rials (e.g., reactants, reagents, additives of best operation can be used to as a goal
and other feeds). The efficiency is first to be possibly achieved by smarter instru-
expressed as a ratio of the energy use per ments or better installations less sensitive
unit mass of product produced (e.g., kJ/ to ambient conditions or smarter controls
kg) or energy produced (kJ/kJ) and then thru procedural automation or state based
ideally in terms of ratio of cost to revenue control as discussed in the in the Sept 2016
by including the cost of energy used (e.g., $ Control Talk Column “Continuous improve-
per kJ) and the value of revenue for prod- ment of continuous processes”.
uct produced (e.g., $ per kg) or energy
produced (e.g., $ per kJ). The kJ of energy The metrics that affect process capacity
and kg and mass are running totals where are more diverse and complicated. Process
the oldest value of mass flow or energy capacity can be affected by feed rates,
multiplied by a time interval between onstream time, startup time, shutdown time,
measurements is replaced in the total by maintenance time, transition time, spectrum
the current value. A deadtime block can of products and their value, recycle, and off
provide the oldest value. The time interval spec product. An increase in off spec prod-
between measurements and the deadtime uct that can be recycled can be taken as a
representative of the time period for the loss in product capacity if the raw material
running total should both be chosen to feed rate is kept the same or taken as a loss
provide a good signal to noise ratio. The in process efficiency if the raw material feed
deadtime block time period should also rate is increased. If the off spec product can
be chosen to help focus on the source of be sold as a lower revenue product, the $
changes in process efficiency. For batch per kg must be correspondingly adjusted.
For batch operations, an increase in batch The metrics that indicate loop performance
end point in terms of kg of product pro- can be classified as load response and set-
duced and a decrease in batch cycle time point response metrics. The load response
including time in-between batches can is often most important in that the desired
translate to an increase in process capac- setpoint response can be achieved for the
ity. If a higher endpoint can be reached best load response by the proper use of
by holding or running the batch longer, PID options. The load response should in
there is a likely increase in process effi- nearly all cases be based on disturbances
ciency assuming a negligible increase in raw that enter as inputs to the process whereas
material but there may be an increase or many academic and model based stud-
decrease in process capacity. The optimum ies are based on disturbances entering
time to end a batch is best determined by in the process output. For self-regulating
looking at the rate of change of product processes where the process deadtime is
formation (batch slope) and if necessary the comparable to or larger than the process
rate of change of raw material and energy time constant, the point of entry does not
use to determine the optimum time to end matter because the intervening process
the batch and move on. A deadtime block time constant does not appreciably slow
is again used to provide a fast update with down input disturbances in the time frame
a good signal to noise ratio to compute the of the PID response (e.g., 2 to 4 deadtimes).
slope of the batch profile and the predic- However, most of the more interesting tem-
tion of batch end point. Of course whether perature and composition control loops in
downstream units for recovery and purifica- my career did not have a negligible process
tion are able to handle an increase in batch time constant and in fact had a near-inte-
capacity and their metrics must be included grating, true integrating or runaway open
in the total picture. For example in etha- loop response.
nol production, a reduction in fermenter
cycle time may not translate to an increase The load metrics are peak error and
in process capacity because of limitations integrated error. The peak error is the max-
in distillation columns downstream or the imum excursion after a load upset. The
dryer for recovery of dried solids byproduct integrated error is most often an absolute
sold as animal feed. For more on the opti- integrated error (IAE) but can be an inte-
mization of batch end points see the Sept grated square error. If the response is non
2012 Control feature article “Getting the oscillatory, the integrated error and IAE
Most Out of your Batch”. are the same. There are also metrics indic-
ative of oscillations such as settling time
and undershoot. The ultimate and practical
limits to peak error are proportional to the to the reset time (no lead) corresponds to a
deadtime and inversely proportional to con- PID structure of Proportional and Derivative
troller gain, respectively. The ultimate and on the Process Variable and Integral action
practical limits to integrated error are pro- on the Error (PD on PV and I on E).
portional to the deadtime squared and the
ratio of controller reset time to controller See the Sept and Oct 2016 Control Talk
gain, respectively. Blogs “PID Options and Solutions - Part
1” and “PID Options and Solutions - Parts
For setpoint metrics, there is the time to 2 and 3” for a discussion of loop metrics
get close to setpoint, which I call rise time, in great detail including when they are
important for process capacity. I am sure important and how to improve them. Also
there is a better name because the metric look at the presentation for the ISA Mentor
must be indicative of the performance for Program WebExs “ISA-Mentor-Program-
an increase or decrease in setpoint. The WebEx-PID-Options-and-Solutions.pdf”.
other setpoint metrics are overshoot,
undershoot and settling time that can My last bit of advice is to ask your spouse
affect process capacity and efficiency. The for metrics on your marriage. Minimizing the
use of a setpoint lead-lag or PID structure deadtime while still having a good signal
that minimizes proportional and derivative to noise ratio is particularly important. For
action on setpoint changes can reduce men, the saying “Happy wife, happy life” I
overshoot, despite using good load distur- think would work the other way as well. I
bance rejection tuning. A setpoint lag equal just need a rhyme.
fact
Reliable, precise temperature transmitter
that aces the test of time
OPTITEMP TT 51 –
technology driven by KROHNE
• 2-wire headmounted temperature transmitter for RTD and TC sensors
• Dual sensor input: sensor drift detection, automatic switchover
to backup sensor
• SmartSense sensor isolation monitoring to detect moisture in the thermowell
• ATEX, IECEx, SIL2
When is an automation
system too slow and too fast?
By Greg McMillan
T
he usual concern is whether an auto- rate and PID execution rate is too slow.
mation system is too slow. There The question is what are the criteria and
are some applications where an priorities in terms of increasing the speed
automation system is disruptive by being of response.
too fast. Here we look at what determines
whether a system should be faster or slower The key to understanding the impact of
and what are the limiting factors and thus slowness is to realize that the minimum
the solution to meeting a speed of response peak error and integrated absolute error
objective. In the process, we will find there are proportional to the deadtime and dead-
are a lot of misconceptions. The good news time squared, respectively. The exception
is that most of corrections needed are is deadtime dominant loops that basically
within the realm of the automation engi- have a peak error equal to the open loop
neer’s responsibility. error (error if the PID is in manual) and thus
an integrated error that is proportional to
The more general case with possible safety deadtime. It is important to realize that
and process performance consequences is this deadtime is not just the process dead-
when the final control element (e.g., control time but a total loop deadtime that is the
valve or variable frequency drive), trans- summation of all the pure delays and the
portation delay, sensor lag(s), transmitter equivalent deadtime from lags in control
damping, signal filtering, wireless update
loop whether in the process, valve, mea- deadtime increasing the actual integrated
surement or controller. error by a factor of thirty six when the new
deadtime is 3 times the original deadtime.
These minimum errors are only achieved by
aggressive tuning seen in the literature but Consequently, how fast automation system
not used in practice because of the inevi- components need to be depends on how
table changes and unknowns concerning much they increase the total loop deadtime.
gains, deadtime, and lags. There is always The components to make the loop faster is
a tradeoff between minimization of errors first chosen based on ease such as decreas-
and robustness. Less aggressive and more ing PID and wireless execution rate, signal
robust tuning while necessary results in filtering and transmitter damping assuming
a greater impact of deadtime in that the these are more than ten percent of total
gain margin (ratio of ultimate gain to PID loop deadtime. Next you need to decrease
gain) and the phase margin (degrees that the largest source of deadtime that may
a process time constant can decrease) take more time and money such as a better
is achieved by setting the tuning to be a thermowell or electrode design, location
greater factor of deadtime. For example, and installation or a more precise and faster
to achieve a gain margin of 6 and a phase valve. The deadtime from PID and wireless
margin of 76 degrees, lambda is set as 3 update rates is about ½ the time between
times the deadtime. updates. The deadtimes from transmitter
damping or sensor lags increase logarithmi-
The actual errors get larger as the tuning cally from about 0.28 to 0.88 times the lag
becomes less aggressive. The actual peak as the ratio of the lag to the largest open
error is inversely proportional to the PID loop time constant decreases from 1 to 0.01.
gain. The actual integrated error is pro- The deadtime from backlash, stiction and
portional to the ratio of the integral time poor sensitivity is the deadband or reso-
(reset time) to PID gain. Consider the use of lution limit divided by the rate of change
lambda integrating process tuning rules for of the controller output. Fortunately,
a near integrating process where lambda deadtime is generally easier and quicker to
is an arrest time. If you triple the deadtime identify than the open loop time constant
used in setting the PID gain and reset to and open loop gain. See the Control Talk
maintain a gain margin of about six and a Blog “Deadtime, the Simple Easy Key to
phase margin of 76 degrees, you decrease Better Control.”
the PID gain by about a factor of two times
the new deadtime and increase the reset For flow and pressure processes, the pro-
time by about a factor of two times the new cess deadtime is often less than one second
making by far the control system compo- four times its ultimate period to reduce res-
nents the largest source of deadtime. For onance and interaction. Hopefully, this is
compressor, liquid pressure and furnace done by making the more important loop
pressure control, the control valve is the faster but if necessary is done by making
largest source of deadtime even when a the less important loops slower. A less rec-
booster is added. Transmitter damping is ognized but very common case of needing
generally the next largest source followed to slow down an automation loop is when
by PID execution rate. it creates a load disturbance to other loops
(e.g., feed rate change). While step changes
There is a common misconception that the are what are analyzed in the literature so far
wireless update time should be less than a as disturbances, in real applications there
fraction (e.g., 1/6) of the response time. For are seldom any step changes due the tuning
the more interesting processes such as tem- of the PID and the response of the valve.
perature and pH, the time constant is much This effect can be approximated by apply-
larger than the deadtime. A well-mixed ing a time constant to the load disturbance
vessel could have a process time constant and realizing that the resulting errors are
that is more than 40 times the process reduced compared to the step disturbance
deadtime. If you use the criteria of 1/6 the by a factor that is one minus the base e to
response time assuming the best case sce- the negative power of lambda divided by
nario of a 63% response time, the increase the disturbance time constant.
in deadtime can be as large as 3 times the
deadtime from the wireless update rate. Overshoot of a temperature or pH setpoint
Fortunately, wireless update rates are never is extremely detrimental to bioreactor
that slow. Another reason not to focus on cell life and productivity. Making the loop
response time is because in integrating response much slower by much less aggres-
processes where there is no steady state, a sive tuning settings and a PID structure of
response time is irrelevant. Integral on Error and Proportional -Deriv-
ative on Process Variable (I on E and PD
The remaining question is when is the auto- on PV) is greatly needed and permitted
mation system too fast? The example that because the load disturbances from cell
most comes to mind is when the faster growth rate or production rate are incred-
system causes greater resonance or inter- ibly slow (effective process time constant
action. You want the most important loops in days). In fact, fast disturbances are the
to be able to see an oscillation from less result of one loop affecting another (e.g.,
important loops whose period is at least pH and dissolved oxygen control).
In dryer control, the difference between The averaged pH setpoint should be cor-
inlet and outlet temperatures that is used rected by a downstream pH loop that is
as the inferential measurement of dryer on a well-mixed vessel that sees a much
moisture is filtered by a large time constant smoother pH on a much narrower region
that is greater than the moisture controller’s of the titration curve. A better solution is
reset time. This is necessary to prevent a signal characterization. The static mixer
spiraling oscillation from positive feedback. controlled variable becomes the abscissa
of the titration curve (reagent demand)
Filters on setpoints are used in loops whose rather than the ordinate (pH). This linear-
setpoint is set by an operator or a valve ization greatly reduces the oscillations from
position controller to change the process the steep portion of the titration curve and
operating point or production rate. This enables a larger PID gain to be used. The
filter can provide synchronization in ratio titration curve must not be very accurate
control of reactant flow maintaining the but must include the effect of absorption
ability of each flow loop to be tuned to deal of carbon dioxide from exposure to air
with supply pressure disturbances and posi- and the change in dissociation constants
tioner sensitivity limits. However, a filter on and consequently actual solution pH with
a secondary lower loop setpoint in cascade temperature not addressed by a standard
control is generally detrimental because it temperature compensator that is simply
slows down the ability of the primary loop addressing the temperature effect in the
to react to disturbances. Nernst equation. You need to be also aware
that the pH of process samples and con-
Finally, more controversial but potentially sequently the shape of the titration curve
useful is a filter on the pH at the outlet of can change due to changes in sample liquid
static mixer for a strong acid and base to phase composition from reaction, evapora-
control in the neutral region. Here the filter tion, absorption and dissolution. The longer
acts to average the inevitable extremely the time is between the sample being taken
large oscillations due to nearly non-existent and titrated, the more problematic are
back mixing and the steep titration curve. these changes.
The result is a happier valve and operator.
50
Years of R
e li a b
ilt y
To learn more about our Dual Input Temperature
Transmitter, go to: www.miinet.com/THZ3TDZ3
1968 - 2018 Or call 800-999-2900
www.ControlGlobal.com
Optimization of
cooling systems
One of the few processes whose efficiency can be doubled
through optimization.
By Béla Lipták
I
n recent months, I discussed some Most cooling systems consist of four heat
non-industrial processes, such as con- transfer steps. Heat is first passed from the
trol of the economy or climate change, process to chilled water, then from chilled
automation of smart cars, the risks of cyber water to Freon, then from Freon to cooling
terrorism, the potentials and limitations of tower water, and finally from cooling tower
artificial intelligence, etc. Now, I’ll return water to ambient air.
to some more traditional topics, focusing
on control and optimization of industrial In each step, heat is transported by fluids that
processes. The first process I’ll discuss is are kept in motion by electric motors (M1 on
optimization of cooling syatems. fan, M2 on compressor, M3 and M4 on pumps,
as shown in Figure 1). The cost of operating the
There are few processes whose efficiency whole system is the sum of the costs of operat-
can be doubled through optimization, but ing these four motors. If we call the total saving
cooling sytems are one of them. The coef- 100%, 60% of that is obtained by optimizing the
ficients of performance of unoptimized compressor, 15% by each of the two pumping
cooling systems are between 2.5 and stations, and 10% by the cooling tower fans.
3.5, meaning that one unit of energy can Savings can be estimated as:
remove 2.5 to 3.5 units of energy from a • 1.0 °F reduction in cooling tower
process. Optimization can double this coef- approach (A0) reduces the operating
ficient to 6 or 7. cost by 1.0%.
• 1.0 °F reduction in cooling tower water approach results in a decrease in the air
temperature (Tctws) reduces the yearly transportation cost. The optimum approach
cost of chiller operation (M2) by 3.0%. (∆T) is determined by collecting the total
• Each °F reduction in chiller ∆T reduces operating cost of both the tower’s fans and
the yearly cost of chiller operation (M2) pumps, and setting the setpoint of TIC-1 in
by 1.5%. Figure 1 to the approach (A0) that corre-
• Yearly pumping costs are reduced by sponds to that minimum cost value on that
about $0.5/GPM for each PSID. curve. TIC-1 is then configured to control
the fan speed.
Optimizing cooling tower fan operation:
The “approach” of a cooling tower is the Optimizing the cooling tower pump oper-
measure of how closely the cooling tower ation: The “range” of the cooling tower is
water temperature (Tctws) approaches the the difference between the water tempera-
wet bulb temperature (Twb), which is the air tures entering and leaving the tower, and
temperature at 100% humidity. An increase the optimum range (R0) is the temperature
in approach results in an increase in water difference (Tctws - Tctwr) that corresponds to
transportation cost and a decrease in the the optimum approach (A0) that results in
operation. The setpoint of TIC-2 in Figure 1 water needs to be pumped), the aim of this
is Tctwr, and TIC-2 manipulates the speeds of optimization strategy is to maximize this T.
the cooling tower pumps.
The T will be maximum when the chilled
Optimizing the chiller compressor opera- water flow rate is the minimum. As the
tion: The yearly operating cost of a chiller is chilled water temperature (Tchws) is already
reduced by about 1.5% for each 1 °F (0.6°C) controlled, this ΔT will be maximized when
reduction in the temperature difference Tchwr is maximum. This goal can be reached
across this heat pump. To minimize this by measuring the opening of the most open
difference, both the chilled water supply chilled water valve in Figure 1 and, if it is not
temperature (Tchws) and the return tempera- yet 90% open, further open it. This increase
ture (Tchwr), must also be maximized. in opening can be obtained by lowering the
P across the process users (lowering the
Figure 1 shows how Tchws is maximized by setpoint of PDIC-1). This lowering of the
the TIC-3 controller, which manipulates the P across the users reduces the yearly pump
speed of the chiller compressor. The set- operating cost (M4) by approximately 50
point of TIC-3 is set by the valve position cents per gpm for each psid.
controller (VPC-1), which keeps increasing
the TIC-3 setpoint until the opening of the The setpoints of the two valve position con-
most open control valve, which is serv- trollers (VPC-1 and VPC-2) will determine if
ing the highest process load, and rises to these adjustments are to occur in sequence
90%. If the opening of even the most open or simultaneously. If both setpoints are
valve is less than 90%, the TIC-3 setpoint adjusted to be the same, simultaneous
is increased, and if the opening exceeds action will result. But, if one adjustment is
90%, its setpoint is reduced. This is called economically more advantageous or safer
load-following control. than the other, the setpoint of the corre-
sponding VPC should be set lower than the
Optimizing the chilled water pumping other.
operation: The combined cost of operat-
ing the chilled water pumps and the chiller This is the case in Figure 1, where, VPC-1 is
compressor itself (M4 + M3) is a function of set at 90% and VPC-2 at 95%. Therefore,
the temperature drop across the evaporator when the most open valve is less than 90%
(Tchwr - Tchws). Because an increase in this T open, VPC-2 will lower the setpoint of PDIC-
decreases compressor operating costs (suc- 1, and VPC-1 will increase the setpoint of
tion pressure rises) while it also decreases TIC-3. When the opening of the most open
pumping costs (the higher the T the less valve exceeds 90% (but has not yet reached
95%), VPC-1 will slowly start lowering the limits are placed on the controller output,
setpoint of TIC-3, while VPC-2 will continue or in case of “selective control” (when the
to lower the setpoint of PDIC-1. controller is not selected for control or when
the output of a cascade master is blocked
If VPC-1 did not lower the TIC setpoint suf- because the loop has been switched to slave
ficiently (or if the lowering was not fast control). To prevent this “integral wind-up,”
enough) and the most-open valve reaches the feedback to the integral mode is taken
95% opening, VPC-2 will take fast correc- from an unblocked signal, such as the valve
tive action by quickly raising the setpoint opening or the measurement signal. This is
of PDIC-1. Thus, no user valve will ever be called “external reset.”
allowed to open fully and go out of control.
Naturally, this will only be the case if the In Figure 1, the load-following controls of
pumps and chillers are sized correctly (to VPC-1 and VPC-2 will be stable only if they’re
be capable of meeting the load) and the integral-only, with their integral time settings
VPCs are correctly tuned. adjusted to about tenfold of the integrals
of their cascade slaves (TIC-3 and PDIC-1),
Tuning, limits and external resets: The which are usually several minutes. This is
integral (also called “reset”) control mode needed to keep the optimization loops stable
integrates past errors, and therefore, its when the selected valve opening signal is
output will saturate if it’s blocked from either cycling or noisy. The high/low limits
reaching the manipulated variable. This is the (HLL-1 and HLL-2) on the setpoints of the
case during startup, when the controller is slave controllers serve to keep them within
switched to manual, when high and/or low safe limits.
New Signal Splitters Offer Easy Set-up SP Series Benefits and Features:
Acromag’s new SP series of isolated signal splitters are ■ Easy configuration via USB with Windows
designed for easier installation and setup. Eight models, software or Agility™ app for Android®
covering a broad range of sensor signals, provide ■ Current, voltage or temperature input
dual 4-20mA outputs proportional to a single current,
voltage or temperature input. Select 2-wire or 4-wire ■ Dual 0/4-20mA, 0-10V, or ±10V outputs.
versions for loop or local power requirements. ■ Input scales independently at each output
A USB connection to a Windows PC or Android device ■ Selectable filtering levels
enables simple, precise configuration of I/O ranges and
a variety of operational settings with free software. ■ Configurable output clamp levels (NAMUR)
Visit www.Acromag.com/SP
TO LEARN MORE
Signal Conditioning & Network I/O Solutions
microBlox® I/O Modules Transmitters, Isolators, Splitters Ethernet, Modbus & Profibus I/O Limit Alarms / Computation
Q
I came across your “Ask the master controller to the flow or pressure
experts” column at www.control- slave controller. Being off-gas from other
global.com, and I’d like to ask a units, the fuel gas composition varies (has
question I’ve had for quite a while regarding H2 and H2S), so during a week, the density
the control of fuel gas to a fired heater. of the fuel gas varies widely between 0.44
and 0.7 Kg/nm. We did some modeling of
This particular unit is a natural-draft heater the effect of gas composition on the master
used in a diesel hydrotreating plant at a temperature controller, and it seemed to
refinery, but I’m actually looking for a gen- be rather small. But beyond that, I can’t
eral answer. The question in itself is very say much.
simple: should we use a flow controller or
a pressure controller to regulate the fuel Juan Florentino, Process control engineer, BP Rotter-
A
are possible choices, what are some guide- As you’ve asked, before answering
lines or considerations to choose between your specific question, I’ll give a gen-
the two? eral description of controlling any
larger fired-gas heater. Figure 1 shows a
What we’re trying to control is the outlet common control system that’s referred to
temperature of the heater. This is the as a parallel, closed-loop control with air
Radiant
section
} LL
Σ TIC
TT
SP
Outlet
excess oxygen trim (AIC) loop, which is
minimizing the amount of excess air that
Firing
otherwise would travel through the burner
rate
(just for the ride), picking up good heat and
• Density compensation slightly improves source pressure is very stable and readings
the accuracy of heat input measurement, are accurate. Generally, the fuel gas pipes
but does not directly measure the heating are small, sometimes less than 2 in. In that
value of the gas. Only calorimeters (CAL) case, flow measurement requires costly
do it. instruments like mass flowmeters. Pressure
measurements are simpler.
For the above reasons, my recommendation
is to use a linear mass flowmeter (5th ed., H.S. Gambhir, Harvindar.S.Gambhir@ril.com
A
surement signal by a calorimeter (5th ed., Based on normal operational
Vol. 2, Ch. 1.27) that measures the heating issues, the answer is that you
value of the gas. could use pressure control, but
it would not be a best option. If you use
In addition, the figure shows the cross-lim- pressure control, you’ll have the follow-
ited parallel metering system for the air-fuel ing issues:
ratio (A/F) controls, which always limits
the fuel flow rate to the available air flow, • The pressure control isn’t really relevant to
so that air is leading on load increase and the flow required to control the heater.
lagging on load decrease. This feature also
protects the system from accidents caused • The pressure control tends to be static
by a fan failure or fuel valve sticking. In when in normal operation, but more diffi-
addition, a minimum fuel flow bypass is pro- cult to handle with upsets.
vided (FCV) to guarantee stable flame, and
a minimum air flow signal is provided (HIC) If you still prefer to do pressure control, you
to match that minimum fuel flow. may need to find the correlation between
flow and pressure in the line vs. the tem-
To prevent interaction between the two perature required in the heater.
valves on the fuel line, make sure the PCV is
far away from the flow control valve. So, the best solution is generally to con-
trol the flow of fuel to the heater. If it’s
Béla Lipták, liptakbela@aol.com compensated to mass flow, it could be
controlled to provide the heater perfor-
A
I’ve always used pressure control mance required.
on cascade successfully. Flow
control also could work well if gas Alejandro Varga, vargaalex@yahoo.com
A
The target temperature (in your ratio control from the furnace feed heat
case, the heater outlet) is influ- inflow calculation.
enced by the amount of the fuel
(mass/unit of time) and its heating value • The two control valves (the PCV at the
(Btu/lb., J/kg, Kcal/kg, etc.), assuming the gas reservoir outlet and the control valve
heater process input (flow and tempera- at the fuel inlet to the heater) should have
ture) is stable. as much gas volume between them as is
practically possible to minimize interac-
Assuming all other parameters (e.g., heating tions between them. This can be achieved
value) are not changing much, controlling by placing them as far apart as possible
the fuel flow gives you direct control of the and/or enlarging the gas line as much
heat input quantity. as possible.
In your case, you’re lucky to have direct In general, I prefer controlling the inflow
measurement of the fuel gas density in of fuel mass to the heater, rather than the
kg/Nm3. The fact that it’s referred to as pressure, because this way you control the
“normal m3” excludes the influence of any amount of heat input. In the past, in fuel oil
variations in the pressure or gas tempera- applications, I had very good results by con-
ture. This gives you the option (which I trolling the pressure on the burners, but it
would have taken, if I were you) to build a was (mainly) because of shortages of other
“calculated point” to have the value of the instruments (like separate flow measurment
fuel gas volumetric flow multipleid by the or fuel density detectors).
fuel gas density, which yields the fuel gas
mass flow. This value should give a very Also, I’ve just come across a very good
good estimation of the total heat input into article in HPC, June 2014, “Automate
the heater. Controlling this quantity should Furnace Controls to Improve Safety
enable you to achieve better control of your and Energy Efficiency” (www.hydrocar-
target temperature. A few more tips: bonprocessing.com/magazine/2014/
june-2014/hpi-focus-energy-efficiency/
• Be sure to use the right size of filter on automate-furnace-controls-to-im-
each of both readings before performing prove-safety-and-energy-efficiency),
the calculation. which doesn’t answer your questions
explicitly, but deals with the problems you
• If the furnace feed flow conditions (flow are facing.
and temperature) are not constant, I
recommend considering feedforward Avihu Hiram, Avihu@HiramEng.com
Q
Your explanation of direct- vs. the use of external reset. These are large
reverse-acting control provided topics that Greg Shinskey explained in
much clarity about fail open (FO) Chapter 2.28 of my handbook (Volume 2,
and fail close (FC) valves, and the controller 4th edition), and I will only briefly discuss
outputs to achieve the desired result (Con- them here.
trol, March ‘14, p. 57, www.controlglobal.
com/articles/2014/controllers-direct-vs-re- Selective control is used whenever the
verse-acting-control). Now, I’m working controlled variables outnumber the
through several problems on override con- manipulated ones (control valves). One
trol, specifically high/low selectors, and application of selective control is the use
would like you to explain these by example of selective overrides. In these systems,
or recommendation for clear, concise read- the override controllers remain inactive
ing material. until their constraint (setpoint) is reached
or exceeded, at which point they take over
Christopher Perumal, christopher_perumal@live.co.za the control of the valve.
A
The various selection techniques A common selective override application is
can not only provide limits or steam flow control with pressure override.
overrides, but can also increase In this case, the control valve is normally
reliability and prevent reset windup by manipulated to keep the steam flow
constant (FIC in Figure 1), but is switched is closed when the bottoms level reaches
over to pressure control if the steam pres- 100%.
sure—to the critical user—reaches the
setpoint of the PIC (soft constraint). If in Another selective override application is
spite of switching from FIC to PIC control, the use of startup and shutdown overrides.
the pressure continues to rise and reaches If a process is too complex for the opera-
the “hard constraint” on the system, alarms tors to manually control during startup or
are actuated or shutdown is triggered. shutdown, automatic override controls are
Override used. Such automatic controls provide over-
Setpoint
Naturally, EF
TIC selective override
FIC control
FY can
PIC be PT rides to protect against violation of several
Master Slave
provided by using multiple override con- constraints. An example is starting up a dis-
TT EF EF
trollers. One example
FT is in boilers, where tillation column, where ramping signals are
we normally control the flow of
FC
combustion usually used to open valves and override
air on carbon monoxide control, but also controls are provided to:
provide overrides to keep opacity, stack
temperature, hydrocarbons and excess • Pinch steam on low base level or on
oxygen within acceptable limits. high pressure differential, which can
cause flooding,
Selective override applications can also
serve to guarantee valve closure. For exam- • Pinch feed valve on high base level, and
ple, when controlling the feed flow to a
distillation column, a bottoms level override • Pinch reflux on low condenser level, and
is provided to make sure that the feed valve increase it on high condenser level.
reliable among multiple sensors or trans- In this configuration, the inputs to the low
mitters. For example, in critical reactor feed selector would be the header pressure
applications or in hostile environments (cor- and the reboiler flow controller outputs,
rosive, dirty, vibrating, high-temperature), and the low selector output would be the
we often assure measurement reliability by lower of the two controllers. In this config-
using median selectors or voting systems uration, when the pressure is higher than
to disregard a failed detector among three the pressure setpoint, the flow controller
sensors. On other processes, such as fixed- controls the valve. If the valve draws too
bed chemical reactors, we can measure the much steam and/or the steam genera-
temperature at several locations and select tion is reduced for some reason and the
the hottest measurement for control. header pressure drops below the pressure
setpoint, the system will start cutting back
Béla Lipták, liptakbela@aol.com the flow to the reboiler (i.e., close the
valve even if the flow requirement is not
A
A common high/low override con- met).
trol situation occurs when trying
to control two variables with only You would use this low selector override
one control valve. Obviously, you can’t con- control configuration when the reboiler
trol both measurements at the same time to flow is less critical and it’s more import-
keep both setpoints constant. One has to be ant to maintain the steam pressure above
the main control variable and the secondary some minimum for the more important
will be under constraint control. process users.
An example of low select override occurs Similar logic is used for high selection.
when controlling steam flow(s) to a reboiler
from a common steam header. As the Simon Lucchini, Simon.Lucchini@Fluor.com
A
opens up and passes more steam. There may I recently implemented a system
be several pieces of equipment on the steam with four masters to decide
header, and there may be a total steam flow which has to take control. If
limitation. Under certain conditions, the you’re interested, I can share it with you.
steam header pressure will get too low. The Don’t hesitate to contact me if you need
constraint control would be to use a pres- more help or more explanations.
sure controller with a setpoint lower than the
normal running pressure of the header. Avihu Hiram, Avihu@HiramEng.com
T
hermowells subjected to flow expe- The wake frequency limit is generally
rience dynamic stresses imposed by the most challenging calculation to
oscillating vortex pressures. These pass, especially for long thermowells or
dynamic stresses can result in vortex-in- high-velocity flows. The limit must be ver-
duced vibration (VIV), the leading cause of ified to ensure that the natural frequency
thermowell stress failures. Emerson is now of the thermowell is safely away from the
offering a patented “twisted square” ther- Strouhal (vortex-shedding) frequency. As
mowell that suppresses VIV, eliminating the these frequencies converge, the thermow-
concern of dynamic stress failure. ell can “lock in” to resonant conditions,
greatly magnifying dynamic stresses
The new design was inspired by the caused by the VIV.
difficulties of specifying conventional
thermowells to avoid issues with VIV. “A conventional thermowell experiences
Most often, stress failures occur on con- lock-in frequencies that multiply stress
ventional thermowells that have not 1,000 times or more,” says Timchan
undergone recommended calculations Bonkat, product manager, temperature,
per ASME PTC 19.3 TW to ensure the Emerson Automation Solutions. “Every
thermowell will withstand fluid forces and material has its own fatigue limit. After
process pressures. enough reps, it will fail.”
The traditional solution to avoid these These situations are becoming more fre-
lock-in regions is to shorten the thermow- quent because process velocities are
ell and/or increase the outer diameter. increasing due to more frequent use of
These changes can result in decreased smaller pipe sizes to save cost on projects.
accuracy or increased response time of
the temperature measurement. Instead of stiffening the thermowell, the
Twisted Square Thermowell desynchro-
“Customers try to standardize on a ther- nizes the vortices in its wake so they are
mowell, so to accommodate the worst not uniformly defined or alternating at a
application, they choose one that’s short consistent phase along the length of the
and fat,” Bonkat says. “They may have thermowell. This dampens the dynamic
to use a larger nozzle size and retrofit stresses from the vortices and suppresses
existing applications. In some of their VIV excitation to a safe level.
applications, the short, fat thermowell
may not reach far enough into the flow for The design reduces resonance stresses
a representative reading.” by more than 90%, so static stresses
dominate. “With a twisted square ther- and it also reduces stocking requirements
mo-well, the static stresses will always because one design can be used in a wide
be the limiting factor. Without resonance variety of applications.
lock-in, the dynamic stresses don’t gen-
erate enough energy to be damaging,” “Above all, it provides an accurate and
Bonkat says. reliable temperature measurement while
preventing thermowell failures that can
“The twisted square simplifies thermowell cause expensive shutdowns, or at worst
calculations, and it works in applications result in leaks and explosions.”
where conventional thermowells don’t
pass ASME PTC 19.3 TW,” Bonkat says. “It For more information, including a link
allows optimum penetration using exist- to an eye-opening video comparing
ing nozzles, and is available in all normal deflections of a conventional thermow-
configurations and materials. It allows ell and a Twisted Square Thermowell
for growth because plants can change under a wide range of flow velocities,
flow rates and other operating conditions see www.emerson.com/en-us/catalog/
without resonant frequencies concerns, rosemount-twisted-square-thermowell.
MillerCoors
optimizes fermentation
CO2 recovery and brewing time insights are provided by pressure readings
from differential pressure level gauges.
By Andrew Klosinski
begins, but it doesn’t always happen at the CO2 production is also not uniform. It fol-
same time with every batch. The standard lows a curve from low to high and then
procedure in the brewery was to wait a spe- back down, but the shape of the curve
cific number of hours before collecting CO2 depends on the recipe, and can change
because there was no useful indication of from batch to batch because the action of
when the yeast had hit its stride. yeast is not always predictable. Moreover,
the CO2 collection and processing system
Since the ERS system allows access to the has its limits and can be overwhelmed, rais-
individual measurements from both sensors, ing backpressure in the headers.
we were able to read the pressure in the
headspace. Having that reading helped us The yeast acts more rapidly in some
realize that CO2 production was beginning batches than in others, and in these
earlier than we thought, so we changed cases, cooling must be applied to retard
the procedure to begin collection when the action. Again, prior to installation of
pressure reached a tipping point, which the ERS system, it was difficult to tell
was often many hours ahead of the nor- when the yeast might force the fermen-
mally scheduled time. With a traditional DP ter past its design working limit and blow
system using impulse/capillary lines, this out a rupture disk. When such an incident
secondary measurement would only have occurs, substantial amounts of CO2 are
been possible by adding another pres- released to atmosphere from the individ-
sure transmitter. ual fermenter and header system until
the situation is discovered and the disk The overall costs related to overpressure
is replaced. and rupture disk problems is substantial,
estimated at $65,000 annually in mainte-
With the ERS system installed, it’s easy to nance and lost CO2, so solving the problem
determine when gas pressure is building was critical to improving profitability with
up in the fermenter headspace (Figure 3). minimal additional cost. The answer was pro-
If CO2 production is increasing more than vided by a simple level transmitter upgrade.
expected, cooling can be increased until it The savings related to pressure issues and
subsides, and vents can be opened if neces- CO2 production were substantial, but a more
sary to prevent blowing a rupture disk. significant opportunity soon emerged.
By Allan Evora
T
he University of North Carolina (UNC) was becoming difficult to find parts for
at Chapel Hill was spending too much the controllers used to control tempera-
time and money supporting an exist- ture and other variables (Figure 1). During
ing control system. That’s unacceptable for maintenance of one particular chamber,
UNC Chapel Hill, which has nearly $1 billion UNC staff couldn’t find a supplier to repair a
in annual research expenditures, and ranks malfunctioning temperature controller. This
eighth in research activity among public and maintainability concern was recognized as
private universities in the United States. a possible campus-wide issue since many of
the chambers were put into service at about
Much of this research is done in about 1,000 the same time.
cold, warm and freezer chambers located
around campus. Some chambers store mil- In addition, needs were expanding, with
lions of dollars worth of products, while researchers requesting tighter, more accu-
others perform critical research in life sci- rate temperature and humidity control,
ence. In many cases, chamber temperatures and the ability to view temperatures in
must be controlled accurately to within 0.5 real time. None of these requests were
°C due to the controlled experiments inside. possible with the old control and mon-
itoring systems. Investigation showed
Many research chambers at UNC were it would be too expensive to upgrade
installed more than 20 years ago, and it the existing proprietary systems, so the
The new controller provides smoother and READY FOR THE FUTURE
Figure 2: Affinity Energy replaced the obsolete
more accurate control than the legacy
controller with an AutomationDirect PLC and
control systems. The controller enclosure HMI due to the open, precise control capabilities
and ease of use of these components.
is built with the PLC and other control
devices (drivers, relays, etc.) as a unit.
Field devices are wired to terminal blocks for several separate Ethernet communica-
included for connections, and the panel tions ports, each running multiple protocols.
is mounted on a wall outside the environ- Do-more PLCs are programmed with
mental chamber. Do-more Designer software, which is avail-
able free online.
The Do-more PLC is fast, has plenty of
memory, and includes multiple commu- Affinity Energy also installed an Auto-
nication options such as serial, USB and mationDirect C-more human machine
embedded Ethernet. It can be configured interface (HMI). This color touchscreen
returning from the evaporator coil in the injecting a portion of the gas leaving the
chamber. This high-pressure refrigerant gas evaporator coil inside the chamber back
condenses into its liquid state, but can only into the inlet of the evaporator coil.
do so if it can release heat.
Adding some of the hot gas into the liquid
As the gas travels through a condenser entering the evaporator coil slows the
coil located outside of the chamber, it cooling process, bringing the tempera-
gives off heat, becoming a high-pres- ture toward a desired point. Since the
sure liquid. This liquid refrigerant liquid line valve is always open, the hot
travels through an orifice and becomes gas valve is modulated at the duty cycle
a low-pressure liquid, which wants to required to reach and maintain a specific
evaporate to a gas, but must absorb heat temperature.
to fuel the transformation. The gas trav-
els through the evaporator coil located Humidity is controlled by heat injection to
inside the chamber, where it can absorb manage how much moisture in the chamber
heat, cooling the room and becoming a condenses on the evaporator coil.
low-pressure gas, which is returned to the
compressor to begin the cycle again. A custom PID control method was needed
to manage both the positive and the
In this application, temperature is controlled negative side of the operation curve.
and maintained by managing the liquid line “Keeping temperatures within a half a
and hot gas valves. The liquid line valve is degree Celsius, with researchers going in
open all the time except during a defrost and out all day long, is not an easy thing,
cycle. As a result, the temperature in the but Affinity Energy’s integrated solution
room would continue to become colder has been running flawlessly,” notes Steve
and colder, reaching temperatures below Hargett, supervisor, facilities maintenance
freezing, were it not for the hot gas valve at UNC Chapel Hill.
The plan is to replace controllers on from the North Carolina State Energy Management
other chambers across the UNC campus, program, and is as a Certified Measurement & Verifi-
using this open design based on Auto- cation Professional (CMVP). He established Affinity