Flooded Condenser Controls:

Principles and
Troubleshooting
Flooded condensers are the prime tower pressure-control methods for total condensers that
generate only liquid products, and although these control methods can be troublesome, a good
understanding of their principles will help achieve improved, trouble-free operations

FIGURE 1. Distillation columns

Henry Z. Kister
are crucial in many facilities, and Fluor Corp.
pressure control within the tower
is of the utmost importance in
ensuring steady operations IN BRIEF
COMMON CONTROL
ARRANGEMENTS

HYDRAULICS OF
FLOODED CONDENSERS

VAPOR PRESSURE
DIFFERENTIALS

OVERCOMING
CHALLENGES

VAPOR TO VAPOR AND

LIQUID TO LIQUID LINES

SURFACE AGITATION

INERT PADDING

DECANTING WATER

NON-CONDENSABLE
GASES

INSUFFICIENT

P
ressure is the most important
variable for controlling distilla-
tion columns (Figure 1) because
pressure affects every aspect of
a distillation system: vaporization, conden-
sation, temperature, volatility and so on.
An unsteady pressure typically results in an
unsteady column.
There are several ways to control tower
pressure, depending on how the tower is
configured. If a tower has an overhead vapor
product, manipulating the vapor flowrate
usually controls pressure. If the tower has no
vapor product (it has a total condenser and
produces only liquid), tower pressure can be
controlled by partially flooding the condenser
and manipulating the liquid level in the con-
denser. Another alternative for either vapor or
SUBCOOLING
liquid products is to manipulate the coolant
flowrate (or temperature) to control the tower AIR CONDENSERS
pressure. Coolant manipulation is popular in VALVE CONFIGURATIONS
refrigerated towers, but is usually avoided in
cooling-water condensers, as it can cause FLOODED DRUM
SCHEME
accelerated fouling and corrosion.
Flooded condenser control is by far the FINAL REMARKS
preferred pressure-control method used
with water-cooled total condensers (those
generating liquid products only). It is also
common with air-cooled total condensers.
In this control method, the condenser area is
partially flooded by condensate. The flooded
tubes do not contact the vapor and perform
little condensation. The column pressure is
controlled by manipulating the flooded area.

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2016 37

 FIGURE 2. These common
flooded-condenser control
arrangements show two dif-
ferent control-valve configura-
tions. In Figure 2a, the control
valve is located in the conden-
sate liquid line, and in Figure
2b, the control valve is placed
in the vapor line
38
PC
Over-
head
from
~
~
P1
Vent
tower
Equalizing
line
Pout
P1 Signal to
LC product
or reflux
valve
Product plus
reflux
a b
Raising the liquid level in the condenser
floods additional tubes, which reduces con-
densation area, thereby raising tower pres-
sure. Conversely, lowering the liquid level in
the condenser exposes more tubes, which
increases the condensation area, and sub-
sequently lowers the column pressure.
The principles of flooded condenser con-
trols were described in literature more than
60 years ago. Chin’s classic paper on dis-
tillation pressure-control describes many of
the principles and good practices [1]. Yet,
these methods continue to be among the
most troublesome distillation controls. A
good understanding of the principles, as well
as learnings from past experiences, are key
for avoiding many of the potential problems
[2]. This article provides an updated and de-
tailed description of the principles of flooded
condenser control, and applies them to ad-
dress many of the most common traps that
can cause operational issues.
Common control arrangements
Although the flooded area performs little
condensation, it serves the vital purpose
of subcooling the condensate. Subcooling
is beneficial when pumping volatile liquids
[3]. Although the subcooling consumes
some heat-transfer area, this area is not
always added in the exchanger design.
Some designers are comfortable to as-
sume that the subcooling area can come
from the overage included in the exchanger
design [3]. Conversely, others prefer to
oversize the condensers by as much as
25% to ensure subcooling, especially when
the subcooling is critical, as in hot-vapor-
bypass schemes [4].
Figure 2 shows two common flooded-
condenser control arrangements, both with
the condenser mounted above the reflux
drum. Figure 2a has the control valve in
CHEMICAL ENGINEERING
PC
Over-
head
from
~
~
P1
tower
Vent
Pin
Equalizing
line
P1 Signal to
LC product
or reflux
valve
Product plus
reflux
the condenser liquid outlet. The required
control valve is small, and should be lo-
cated as close to the reflux drum as pos-
sible to maximize static head when the
condensate enters at the top of the drum
[1]. For condensate entrance at the bottom
of the reflux drum (as seen in Figure 3a),
the valve should be located at the lowest
horizontal leg. This method is simple and
linear, and maintains the same pressure in
the column and in the drum. It is therefore
often favored [1, 5].
This control method requires that a pres-
sure-equalizing line is included [1, 5, 6, 7].
Without this line, the pressure in the reflux
accumulator will be unsteady. A smaller
equalizing line is required when the sub-
cooled liquid is introduced near the bottom
of the drum, as shown in Figure 2a.
Figure 2b shows a flooded condenser
scheme similar to that in Figure 2a, but with
the control valve located at the condenser
vapor inlet. Similar to the method in Figure
2a, the condenser liquid outlet line must
enter near the bottom of the reflux drum,
and a pressure-equalizing line is required.
Placing the control valve in the vapor inlet
(Figure 2b) renders the condensation pres-
sure lower than when the valve is in the con-
densate outlet (Figure 2a), resulting in the
requirement of additional condenser surface
area. If no additional area is provided, tower
pressure must be raised, which increases
energy consumption. The required vapor-
control valve is large and may be expensive
with large overhead lines.
Figure 3 shows three additional flooded-
condenser control schemes, all containing
a control valve in the condenser vapor by-
pass. Figures 3a and 3b are analogous to
Figures 2a and 2b; the only difference being
the addition of the bypass control valve.
This control valve helps overcome the pres-
WWW.CHEMENGONLINE.COM JANUARY 2016
 PT PT FIGURE 3. Three flooded-
Over-
head
from
P1 ~
~
Overhead
from tower
~
~
P1 condenser control schemes
with valves in the vapor by-
Vent Pin pass are shown: Figure 3a
tower Pbypass PC Pbypass
DPC DPC places the pressure-control
Vent
Pcond Pcond valve in the condensate liq-
uid line; Figure 3b places the
PC H PT PT pressure-control valve in the
Signal to H
Signal to
vapor line; and in Figure 3c,
P2 P2 the pressure-control valve is
LC product LC product
or reflux or reflux in the hot-vapor bypass
valve valve

Product plus Product plus

Pout reflux reflux
a b

sure drop in the condenser, especially when PT

the liquid head is small.
Figure 3c shows the hot-vapor-bypass
pressure-control method. Here, the location
of the pressure-control valve is switched
from the condenser route in Figures 2, 3a
and 3b to the hot-vapor bypass. The con-
denser is at ground level, typically 10–20 ft
below the reflux drum (which may be hori-
zontal or vertical). The condensate must be
subcooled so that the liquid surface in the
drum is colder than in the condenser. While
the pressure in the condenser is equal to
the vapor pressure of the condensing vapor,
the drum pressure is the vapor pressure of
the cooler liquid surface in the drum. The
difference in vapor pressures lifts conden-
sate from the condenser into the drum. To
reduce column pressure, the valve is throt-
tled, reducing the hot vapor supply to the
drum, and the drum’s liquid surface cools.
The colder surface has less vapor pressure.
This increases the pressure difference be-
tween the condenser and the drum, which
in turn sucks liquid from the condenser into
the drum. This exposes additional tubes in
the condenser, and increases the conden-
sation rate, which lowers column pressure.
The hot-vapor-bypass arrangement per-
mits the condensers to be mounted at
ground level instead of on a platform above
the reflux drum. Locating large cooling-wa-
ter condensers at ground level eliminates
the requirement for a massive condenser-
support structure, and there is also no need
to pipe cooling water to high elevations.
This provides easy access for maintenance,
the piping is simple, the control valve is
small, and the response is fast [3, 8]. These
advantages can translate into considerable
savings in steelwork, platforms, trolleys and
maintenance. These savings can be major
in large installations, especially where a bat-
tery of condensers rather than a single ex-
changer is used. However, this method suf-
fers from many potential issues, which are
described in detail in Ref. 9.
P1
From
tower
Pbypass PC
Signal to
P2
LC product
or reflux
valve
Vent h
Pcond
Product plus
reflux
c
Hydraulics of flooded condensers
Figure 3a shows a very common arrange-
ment. The condenser is elevated above the
drum, with the condensate descending into
the drum due to gravity. The column pressure-
control valve is in the condensate line from the
condenser to the drum, which enters the drum
below the liquid level. Up to this point, this
scheme is the same as Figure 2a. The differ-
ence is that in Figure 3a, there is also a control
valve in the condenser bypass. The presence
of this control valve renders the pressure in the
drum lower than at the condenser, which intro-
duces a vapor-pressure effect. Assuming neg-
ligible line pressure losses, a pressure balance
on the condenser gives Equation (1) below.
P1 – P2 = –H + ∆Pcond + ∆Pout (1)
The variables in Equation (1) are defined
as follows:
• P1 is the pressure at the junction between
the vapor line to the condenser and the
condenser bypass in psia
• P2 is the pressure at the vapor space
inside the reflux drum in psia
• H is the head differential between the
condenser liquid level and the reflux-drum
liquid level in psi
• ∆Pcond is the condenser pressure drop in psi
• ∆Pout is the pressure drop across the
condensate outlet control valve in psi
The density used to calculate the head dif-
ferential H is the difference between the liquid

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2016 39

 and vapor densities, due to the pres-
ence of a static leg of vapor in the by-
pass line. The vapor density is based on
the drum pressure and the drum vapor-
space temperature. The liquid density is
best approximated as the density of the
subcooled liquid leaving the condenser.
Without a control valve in the bypass
(Figure 2a), then P1 = P2, and Equation
(1) becomes Equation (2).
H = ∆Pcond + ∆Pout (2)
Equation (2) states that for the Fig-
ure 2a configuration, the head required
to drain the condenser must be high
enough to overcome the condenser
pressure drop plus the pressure drop of
the condenser outlet valve. This condi-
tion may not be satisfied when the con-
denser pressure drop (∆Pcond) is high,
or the available liquid head H is low.
Adding the valve in the bypass (convert-
ing the Figure 2a scheme to Figure 3a)
changes the difference between P1 and
P2 to ∆Pbypass, giving Equation (3).
P1 – P2 = ∆Pbypass (3)
Combining Equations (1) and (3) gives
the scheme shown in Figure 3a, repre-
sented by Equation (4).
H + ∆Pbypass = ∆Pcond + ∆Pout (4)
Therefore, the pressure drop across
the bypass valve helps the gravity head
push the liquid from the condenser into
the drum. For the arrangement in Figure
3b, the pressure drop at the condenser
inlet ∆Pin (psi) replaces the pressure drop
at the condenser outlet ∆Pout, and Equa-
tion (4) becomes Equation (5) below.
H + ∆Pbypass = ∆Pcond + ∆Pin (5)
For the arrangement in Figure 3c,
there are no control valves at the con-
denser inlet or outlet, and the liquid
head h (psi) is the difference between
the reflux-drum liquid level and the liq-
uid level in the condenser, meaning that
h = –H. The value of H in Equation (4) is
negative, as the liquid level in the reflux
drum is at a higher elevation than that
of the condenser. Equation (4) then be-
comes Equation (6).
∆Pbypass = ∆Pcond + h (6)
40 CHEMICAL ENGINEERING
Vapor pressure differentials
As long as there is no hydraulic restric-
tion in the condenser bypass line (Figure
2a), and the pressure drop of the con-
denser inlet line and at the condenser
entry is low, the pressure is the same at
the condenser as at the reflux drum. In
this case, the vapor-pressure differential
between the condenser surface and the
drum’s liquid surface is zero. In all other
situations, the vapor-pressure differential
plays a role, often a major one.
The pressure differences between the
condenser and the drum in Figures 3a,
3b and 3c introduce vapor-pressure ef-
fects. In Figures 3a and 3c, P1 – ∆Pcond
is the vapor pressure of the liquid sur-
face in the condenser, assuming most
of the condenser pressure drop is near
the condenser inlet — usually a reason-
able assumption for total condensers. In
Figures 2b and 3b, the vapor pressure
of the liquid surface in the condenser is
P1 – ∆Pin – ∆Pcond. In all the schemes in
Figures 2 and 3, P2 is the vapor pressure
at the reflux-drum surface.
In the schemes in Figures 2 and 3, the
hot vapor provided by the bypass con-
denses onto the drum liquid surface,
which keeps the surface hot. At steady
state, the hot bypass introduces sufficient
vapor to maintain the temperature of the
drum’s liquid surface at the value corre-
sponding to the desired vapor pressure
P2. Heat flows from the hot liquid surface
to the subcooled liquid underneath, and
there are also atmospheric heat losses
from the vapor space of the drum. These
heat flows must be matched by con-
densing the hot vapor from the bypass.
As long as the drum surface remains
steady, most of the heat flow from the
surface to the subcooled bulk liquid is
by conduction. Since process liquids are
good thermal insulators, the conduction
heat transfer from the hot surface to the
subcooled liquid is small. In reality, con-
vection and some bulk movement raise
the heat transfer from the surface to the
subcooled bulk liquid, but even account-
ing for these, the bypass vapor flowrate
can easily match the heat demand at the
drum liquid surface.
Ref. 9 details the heat balance for the
drum in relation to the scheme in Figure
3c. The heat-balance discussion pre-
sented there also fully applies to other
flooded-condenser control schemes
(Figures 2a, 2b, 3a and 3b).
WWW.CHEMENGONLINE.COM JANUARY 2016
 The vapor-pressure effects become of ut-
most importance in the hot-vapor-bypass
scheme with submerged condensers (Figure
3c). In this scheme, the vapor-pressure differ-
ences directly determine the flooded height
in the condenser. In the elevated-condenser
schemes (Figures 3a and 3b), the condenser
inlet or outlet valve directly determines the
flooded height, with the bypass mainly used
to provide sufficient pressure drop, per
Equations (4) and (5).
With the hot-vapor-bypass scheme (Figure
3c), the vapor-pressure difference provides
the driving head that pumps the conden-
sate liquid from the condenser into the reflux
drum as described in Equation (6). Equation
(3) can be combined with Equation (6) to
give Equation (7):
P1 – P2 – ∆Pcond = h (7)
Equation (7) shows that the vapor-pressure
difference is balanced by the liquid head lift. If
higher pressure is required in the tower, there
is a need to flood more area in the condenser.
This requires reducing the liquid head lift h.
To achieve this, the vapor-pressure differen-
tial P1 – P2 – ∆Pcond must be reduced. With
P1 and ∆Pcond constant, this is achieved by
opening the hot-vapor bypass to raise P2.
Opening the hot-vapor bypass heats up the
liquid surface in the drum, which raises the
vapor pressure P2. Conversely, to reduce col-
umn pressure, there is a need to lower the
liquid level in the condenser, which raises the
Most importantly, liquid from the condenser must enter the reflux drum
near the bottom of the drum, well below the liquid surface
liquid head lift. This is achieved by closing the
bypass valve. This cools the surface in the
drum and lowers P2. The larger P1 – P2 dif-
ference sucks liquid from the condenser into
the drum, thus exposing more condenser
area for condensation. These mechanisms
are described in detail in Ref. 9.
OVERCOMING CHALLENGES
Correct configuration is mandatory for the
success of all flooded condenser schemes,
due to their challenging nature.
Vapor to vapor and liquid to liquid lines
Bypass vapor must enter the vapor space of
the reflux drum. The bypass should be free
of pockets where liquid can accumulate, and
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
any horizontal runs should drain into the re-
flux drum. The author is familiar with cases
where a pocket of liquid in the hot-vapor-
bypass line in Figure 3c led to severe oscil-
lations and column pressure swings. Most
importantly, liquid from the condenser must
enter the reflux drum near the bottom of the
drum, well below the liquid surface. This is
imperative with the Figure 3c configuration,
and also highly recommended with the other
schemes. The rule is “vapor to vapor, liquid
to liquid.”
If the liquid enters at the bottom of the
drum with an upward momentum (as shown
in Figures 3a, 3b and 3c), a horizontal baffle
should be added above the inlet to spread
the momentum of the incoming liquid jet. As
reported in Refs. 9 and 10, liquid jets at ve-
locities of a few feet per second can easily
penetrate through several feet of drum liquid,
bringing a variable amount of subcooled liq-
uid to the drum liquid surface, disturbing the
surface. In some cases, such disturbances
have caused a massive amount of liquid to
be suddenly sucked from the condenser into
the drum [9, 10].
Figure 4a depicts a case in which viola-
tion of this practice led to severe pressure
fluctuations, an inability to maintain column
pressure and a capacity bottleneck [9, 11].
In this scheme, subcooled liquid mixed
with vapor at its dewpoint, and vapor col-
lapse occurred at the site of mixing. The
rate of vapor collapse varied with changes
in subcooling, overhead temperature and
condensation rate. Variation of this col-
lapse rate induced pressure fluctuations
and hammering.
The green piping in Figure 4b shows the
piping modification that eliminated the prob-
lem. The liquid and vapor lines were sepa-
rated, and the vapor line was altered so that
it introduced vapor into the top of the reflux
drum. After these changes were made, the
tower pressure no longer fluctuated, and the
problem was completely solved.
This case is one example of a violation
of the “vapor to vapor, liquid to liquid” rule
described above, and is the most common
cause of poor performance with hot-vapor-
bypass schemes. A number of these cases
have been reported in literature [5, 6, 11, 12].
JANUARY 2016 41
 PC
Reprinted courtesy of AIChE decrease to as low as the vapor pressure of
the liquid at the subcooled temperature. The
author is familiar with situations where this
pressure loss pulled vacuum inside the reflux
drum. In other cases, drum pressure fluctu-
LC ated, sometimes wildly. In some other situa-
Tower tions where the bypass was large, the liquid
almost entirely lost its subcooling, causing
M

cavitation of the reflux pump. Splashing sub-

M

CW CW cooled liquid onto the drum surface can also

lead to the generation of static electricity.
Four
FC The higher the difference between the bub-
condensers
ble point and the subcooled temperature,
a
the more aggravated these issues become.
PC
Configurations where the liquid line enters
the drum liquid also have the advantage
of providing a better seal to the condenser
and preventing vapor from blowing through
the condenser [3].
LC A common design practice is to introduce
Tower the liquid from the top of the drum via a slot-
ted or perforated pipe, so that most of the
M

liquid is introduced below the drum liquid

M

CW CW level, but some is splashed onto the liquid

FIGURE 4. These diagrams
show correct and incorrect
hot-vapor-bypass hookups.
Figure 4a illustrates a com-
mon mistake that leads to
pressure fluctuations. Figure
4b shows modifications (in
green) that correct the issues
in Figure 4a, and provide
sound pressure control [9]
42
FC
b near the bottom of the drum, and has been
Any subcooled liquid streams entering the
drum must also enter at or near the bottom
of the drum. In one case, subcooled liquid
entered the drum vapor space (presum-
ably due to unflooding of the liquid inlet)
[6]. The vapor space was 100°F hotter, and
rapid condensation sucked the liquid leg
between the drum and condenser into the
drum in seconds.
With the systems in Figures 2a, 3a and
3b, there is some debate in the literature
whether the liquid should be introduced
into the vapor space or into the liquid at (or
at least near) the bottom of the drum. Ref.
1 recommends liquid entry above the liq-
uid level so that drum level does not affect
the condenser level.
The author and others strongly prefer that
the subcooled liquid enters at the bottom
of the drum [8, 13]. Introducing subcooled
liquid above the liquid level is likely to cause
vapor collapse onto the cold liquid. This in
turn may result in pressure fluctuations and
possible hammering. Further, introducing the
subcooled liquid onto the drum liquid sur-
face drops the vapor pressure in the drum
by a large amount, raising the demand of
hot vapor from the bypass tremendously,
and often overwhelming the capacity of the
bypass. Unless a much larger bypass is
available, the pressure inside the drum can
CHEMICAL ENGINEERING
surface. This method is better than introduc-
ing all the liquid into the vapor space, but
is not as good as introducing all the liquid
troublesome. The larger the opening that
discharges liquid into the vapor space, the
more troublesome this method is likely to be,
especially in situations with a high degree
of subcooling, such as during cold winter
nights or low-rate operation. The issues are
identical to those described in the previous
paragraph. In the author’s experience, severe
hammering has occurred when highly sub-
cooled liquid was introduced from the top of
the drum with perforated or slotted dip pipes.
The hammering ceased after the slots in the
vapor space were blocked.
Surface agitation
Operation may be troublesome if the drum’s
liquid surface is agitated. Surface agitation is a
particular concern with the hot-vapor-bypass
scheme (Figure 3c), but the other schemes
are not immune. Such agitation stirs up sub-
cooled liquid and brings it to the hot liquid
surface in the drum, causing fluctuations in
the drum pressure. The source of agitation
may be from impingement of a high-velocity
hot vapor jet on the liquid surface, due to up-
ward-directed subcooled liquid jets reaching
the liquid surface, as well as other causes.
Control instability has been reported when
surface agitation occurred due to strong ex-
ternal vibrations on the reflux drum platform
[14]. Agitation of the liquid surface can often
be avoided by judicious baffling [9].
WWW.CHEMENGONLINE.COM JANUARY 2016
Inert padding
Instability due to surface disturbances or
agitation may be alleviated, even mitigated,
by padding the drum with non-condens-
able inert gases. A source of inerts, such
as nitrogen or fuel gas, is connected to the
vapor space of the drum. The drum pres-
sure is controlled by adding or venting the
inerts. The drum pressure is no longer the
vapor pressure of the liquid, but now equals
the sum of the vapor pressure (VP) of the
drum liquid and the inerts partial pressure,
as shown in Equation (8). The box on p. 44
presents a practical calculation example of
the effects of inerts in a tower.
P2 = VPdrum liquid + Pinerts (8)
With total condensers, inert padding is
usually implemented during operation as
a temporary solution to alleviate instability,
especially since the inerts can be quite ex-
pensive. The vented inerts contain vapor-
ized product. In the calculation example on
p. 44, assuming ideality and equilibrium,
10% (3.1 psia/30 psia) of the vent gas at
80°F on a molar basis will be hexane. On
a weight basis, the hexane fraction of the
vent gas is even higher due to the low mo-
lecular weight of the inerts — 26% hexane
on a weight basis for nitrogen padding.
This vented product is likely to be lost, and
may increase flaring or emissions. The in-
erts, even nitrogen and fuel gas, can be
absorbed into the product, and can later
increase pressure in downstream equip-
ment, resulting in more product loss and
flaring downstream. To avoid inconsistent
transitions from the inert-addition mode to
the venting mode, there is often a pressure
range in which inerts are added simultane-
ously with venting, which compounds the
previously described issues [15]. In order
to maximize product recovery and minimize
emissions and flaring, some experts rec-
ommend against using inert padding with
total condensers, other than as a temporary
solution [15]. Diagnosing the cause of, and
eliminating the surface instability, is usually
a preferred longterm solution, especially
with volatile products.
Decanting water
If the reflux drum is used to decant small
quantities of free water from condensed
hydrocarbons or other water-insoluble or-
ganics, the entry point of the condensate
liquid (and other subcooled liquid streams
that may contain free water) should be lo-
cated within the drum opposite to the end at
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
PC
Overhead
from tower ~
~
Vent
Signal to
LC product
or reflux
valve
ILC Interface
Water Product plus
reflux
which the liquid product and reflux are with- FIGURE 5. The flooded-
drawn. The water-removal boot should be condenser control scheme
from Figure 2a is illustrated
just upstream of the point where the reflux here with a water decanting
and product streams are withdrawn [16], configuration
as illustrated in Figure 5 for the control sys-
tem from Figure 2a. In many cases, a short
standpipe (about 6 to 12 in. tall) or judicious
baffling are used as additional measures to
keep water out of the reflux and product
draw [16], but these additional measures
may lead to water accumulation in the drum.
Also, corrosion is possible when the interface
level controller in the boot malfunctions, and
potentially acidic water is not adequately re-
moved from the drum. Ref. 17 describes a
related experience.
Non-condensable gases
Flooded condenser schemes are suitable
only for total condensers, although some
less satisfactory variations are also available
for partial condensers [5]. The schemes in
Figures 2 and 3 can handle small amounts
non-condensable gases, such as those in-
troduced during startups or upstream up-
sets. To handle these non-condensables,
vents are required on the condenser and the
drum. The condenser vents can be directed
to the vapor space of the drum, to an up-
stream unit or elsewhere. The drum vents
should be board-operated, and if frequent
venting is anticipated, the condenser vents
should also be board-operated. In one case,
a debutanizer flooded-condenser system
experienced frequent high pressure, insta-
bility and flaring due to the breakthrough of
non-condensables from an upstream tower
that had control issues [18]. The problem
JANUARY 2016 43
 THE IMPACT OF INERT PADDING
C
onsider a tower making hexane top product at 35 psia and 210ºF. The over-
head vapor is totally condensed, and is subcooled to 80ºF before entering
the reflux accumulator at 30 psia. With no inerts, the drum liquid surface will
be at about 200ºF to match the vapor pressure of hexane at 30 psia. A disturbance
that lifts 5% of the subcooled liquid to the surface will cool the surface to 194ºF (0.05
× 80ºF + 0.95 × 200ºF), which will in turn drop the drum pressure by 3 psi — quite a
large pressure swing. In contrast, with inerts filling the drum vapor space, the drum
surface can be as cool as the subcooled temperature of 80ºF. At this temperature,
the vapor pressure of hexane is 3.1 psia, with the partial pressure of the inerrts mak-
ing up the remaining drum pressure. A 6ºF drop in surface temperature will lower the
hexane vapor pressure to 2.7 psia, causing only a small change of 0.4 psia to the
drum pressure.
was mitigated by adding a manual board-
operated condenser vent that was opened
upon high pressure and vented to an
upstream system.
Insufficient subcooling
Flooded-condenser control methods pro-
duce subcooled reflux and product. This
subcooling is beneficial in avoiding net
positive suction head (NPSH) issues in the
pump or flashing problems in the reflux or
product line. Such flashing can lead to in-
stability, poor reflux distribution at the tower
inlet, slug flow and even hammering [19].
The instability may be particularly severe
with the flooded drum method (discussed
below), due to its potential interaction with
the column pressure control. However,
other flooded condenser methods can also
exhibit such issues.
Subcooling is diminished when the con-
denser nears its maximum capacity. This
may be the natural maximum limit, or can
be caused by fouling, non-condensable ac-
cumulation, condenser drainage or other is-
sues. Subcooling is also diminished when
liquid is splashed onto the surface of the
drum, as discussed earlier. Finally, many
advanced controls use pressure minimiza-
tion strategies, such as Shinskey’s floating
pressure control [19, 20]. These strategies
reduce tower pressure during periods of fa-
vorable ambient temperatures to conserve
energy, and in high-pressure towers (greater
than 150 psia), can also maximize capac-
ity. The pressure reduction brings the con-
denser closer to its limit, and by doing so,
minimizes subcooling.
Issues with insufficient subcooling can
be avoided, as long as the condenser is
not near its capacity limit. The keys are to
avoid splashing the condensate liquid onto
the drum surface, minimize condenser foul-
ing, properly vent non-condensables from
the condenser and adequately monitor the
subcooling. Some override control, or sim-
44 CHEMICAL ENGINEERING
ply an operator advisory to limit the pressure
minimization upon low subcooling, may also
be beneficial.
Air condensers
Air condensers are elevated above the re-
flux drum, so the only compatible flooded
condenser schemes are those for condens-
ers mounted above the drum (as seen in
Figures 2, 3a and 3b, as well as Figure 7,
discussed further below). The hot-vapor-
❏ bypass scheme (Figure 3c) requires mount-
ing the condenser below the drum, and is
incompatible with air condensers. In one
tower, overhead vapor was condensed in
an elevated air condenser followed by a
ground-level cooling water condenser. The
Figure 3c scheme worked well when the
liquid level was in the water condenser, but
became unstable on cold days when the air
condenser supplied the total condensation
duty and the liquid level climbed into the air
condenser. The solution was to reduce the
air condenser duty by shutting off fans and
closing louvers so the liquid level remained in
the water condenser [9].
An air condenser has a small ratio of
height to width. Any change in liquid level,
even as small as 1 in., may result in an en-
tire tube row being covered or uncovered.
Typically, an air condenser will have very few
rows (well below 10), so covering or uncov-
ering one results in a bump in heat transfer.
It is common to slightly slope several bottom
tube rows, or all rows, towards the outlet so
that the movement of liquid level up or down
the tubes is smoother [4].
Valve configurations
There are unique issues associated with the
various valve configurations shown in Fig-
ures 2 and 3. The following sections detail
these issues and provide some guidance for
avoiding them.
Valve in the condenser vapor inlet line
(Figure 2b and 3b). As mentioned earlier,
this method places a backpressure valve
in the overhead vapor line, thereby re-
ducing condenser temperature difference
and capacity.
To minimize pressure drop, the overhead
valve is often designed for a small pressure
drop when fully open. This often leads to
valve oversizing. When oversized, the valve
operates barely open during winter and
cold spells. In this situation, very small valve
movements cause large fluctuations in tower
pressure. In one case, it was necessary to
throttle a manual valve upstream in the line
WWW.CHEMENGONLINE.COM JANUARY 2016
 PT
to force the control valve to operate close to
its half-open position [16].
This method is prone to liquid hammering
Overhead
from tower
~
~
Vent
if the valve closes excessively. In one case,
the valve closed fully under some startup
conditions [21]. Vapor downstream of the
valve rapidly condensed, causing liquid to be
PC PC
rapidly drawn from the reflux drum, which in
turn generated a liquid hammer that shook
the whole unit. The problem was solved
Signal to
by changing the valve so that it would not LC product
fully close [21]. or reflux
valve
Valves in both the condenser and bypass
lines (Figures 3a and 3b). The addition of Product plus
the second valve generates potential interac- reflux
a
tion between the loops controlling pressure
and differential pressure. There is also the PT
question of which variable should be used to
Over-
head
P1
~
control the bypass. Friedman’s work in Ref. from ~

22 specifically addresses these questions.
Friedman advocates controlling the by-
pass using a separate drum-pressure con-
troller (Figure 6a), rather than the differential
pressure controller in Figures 3a and 3b.
Ref. 1 reports one successful case with the
system shown in Figure 6a where the liquid
head was small and the valve in the bypass
line was needed.
The Figure 6a scheme is uncommon, but
the author is familiar with troublesome inter-
action between the pressure controllers in this
scheme. This interaction is discussed in Refs.
8 and 22. Ref. 22 states that in this interaction,
the two control loops help each other. Both
references also state that the key for success
with this scheme is to tune the drum pres-
sure fast and the column pressure slow, much
like a level controller. However, as previously
stated, tower pressure is the most important
column-control variable. It therefore needs to
be tuned fast so it does not wander. Unlike
pressure, level does not affect many variables,
and as long as it stays within limits, it can move
slower and be allowed to drift. Therefore,
the Figure 6a scheme is not recommended
by the author.
Another issue with using a separate drum
pressure controller is that every time the set-
point is changed on the tower pressure con-
troller, the same change must be made on the
reflux-drum pressure controller [8]. The reflux-
drum pressure setpoint needs to be lower
than the tower pressure setpoint, making the
scheme prone to major upsets if operator
error occurs. To overcome this issue, Ref. 8
proposes an advanced control that subtracts
an appropriate bias from the tower setpoint
to provide the setpoint for the reflux-drum
pressure controller.
~
~
tower Vent
DPC
Pressure
setpoint
PC
PT
P2 Signal to
LC product
or reflux
valve
Product plus
reflux
Pout
b
A widely preferred alternative to the dual FIGURE 6. These schemes
pressure-controller scheme in Figure 6a is present alternatives to dif-
ferential pressure control
the differential-pressure (dP) control scheme on the vapor bypass. Figure
in Figure 3a and 3b. Friedman notes that 6a shows a separate pres-
with this scheme, the two controllers tend sure controller on the reflux
drum, which is not favored
to fight each other. Upon an increase in col- by the author. Figure 6b uses
umn pressure, the condenser outlet valve the pressure setpoint in the
(Figure 3a) opens, lowering the liquid level in differential pressure control-
the condenser, while the dP controller opens ler, and alleviates concerns
regarding negative controller
the bypass valve, which raises the liquid level interactions
in the condenser. While there is some de-
bate in industry about the operability of this
control scheme, it is accepted that if both
controllers are tuned fast, there may be an
unfavorable interaction.
A simple solution, practiced by many of
those that reported the scheme to be trou-
blesome, is to tune the dP valve slow or to
place the dP valve in manual mode. These
solutions have been implemented success-
fully to overcome the controller interactions.
It is important to keep in mind that the main
objective of the dP valve is to provide a re-
striction that will overcome the pressure
drops on the righthand sides of Equations
(4) and (5), so there is no need for tight con-
trol of the bypass pressure drop. One must

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2016 45

 simply ensure that adequate resistance
is present.
In Figure 6b, Friedman presents a more el-
egant, albeit less widely practiced, solution.
Instead of controlling the actual dP, the dP is
set as the difference between the setpoint on
the pressure controller and the drum actual
pressure. This scheme has the strength of
Figure 3a, which maintains a set dP, without
getting into the negative interactions. Also,
this configuration permits fast tuning of the
pressure controller, which is essential. As in
the Figures 3a and 3b schemes, the pres-
sure should be tuned fast and the dP slower.
Remember that the dP’s only purpose is to
provide sufficient pressure drop across the
condensate valve.
To mitigate potential interactions between the drum and the condenser
liquid levels, the pressure controller should be tuned much tighter than
the drum level controller
Hot-vapor-bypass controls (Figure 3c).
In the hot-vapor-bypass configuration with
submerged condensers, there are numer-
ous issues, which are covered in great detail
in Ref. 9. The following are the key issues to
keep in mind with this scheme:
• Correct piping is mandatory for the suc-
cess of the hot-vapor-bypass control
method. As described earlier, the bypass
vapor must enter the vapor space of the
reflux drum (Figure 3c or Figure 4b). The
bypass should be free of pockets where
liquid can accumulate, and any horizontal
runs should drain into the reflux drum.
Most importantly, liquid from the con-
denser, as well as any other subcooled
liquid streams, such as the reflux-pump
minimum-flow recycle stream, must enter
near the bottom of the reflux drum. Many
experiences have been reported in which
incorrect piping led to instability, poor
control and hammering [4–6, 11, 12, 23]
• As previously described, operation may
be troublesome and unstable if the drum’s
liquid surface is agitated [1, 6, 9, 10]
• A sudden reduction in drum pressure
can rapidly suck the liquid out of the
condenser, causing a major upset [9,10].
There is also the possibility of U-tube
oscillations [5, 24, 25]. Both issues can
be mitigated by adding a throttling valve
in the liquid leg between the condenser
and the drum
46 CHEMICAL ENGINEERING
• There is a potential for interaction be-
tween the drum and the condenser liquid
levels [5, 9, 10, 20, 24, 25]. To mitigate
the interaction, the pressure controller
should be tuned much tighter than the
drum level controller [20, 24]. This can be
an issue if the reflux drum is small, and
the level controller needs to be tuned fast
to avoid overflow or loss of level. Al-
though this scenario is quite uncommon,
the author has experienced it, and Ref.
26 reports an additional case where
this occurred
• Because of the liquid leg between the
condenser and the drum, non-condens-
ables accumulate in the condenser and
need venting from their accumulation
points. If a vent line is absent, instability
and capacity bottlenecks may result [27]
• Leakage of vapor through the bypass
valve at the closed position can sub-
stantially reduce condenser capacity. In
one case, closing a manual valve in the
hot-vapor bypass increased condenser
capacity by 50% [28]
• Undersizing the bypass control valve
may lead to an inability to maintain the
tower pressure during cold winter days
when the drum is not insulated. In one
situation, poor pressure control due to
undersizing was improved by installing a
throttling valve in the liquid line from the
condenser to the drum [29]
• In some cases, the hot-vapor-bypass
control valve is manipulated by the drum
pressure instead of the tower pressure
[29]. Dynamically, this control scheme is
inferior because the vapor volume in the
drum is much smaller than in the tower
and more variable in response to
ambient changes
• The reflux-drum vapor space may require
insulation to minimize interference from
rain and snow [1, 5, 6, 9]. This issue is
reported to be more pronounced with
narrow boiling-range mixtures [1, 6],
and at high pressure [6], where small
temperature changes have a large effect
on the split of overhead flow between
the condenser and the bypass. At the
WWW.CHEMENGONLINE.COM JANUARY 2016

other extreme, the incidence of Rayleigh
fractionation with wide-boiling mixtures
(where heavy components condense
out without combining with the remain-
ing mixture), can also interfere with this
control system [3]
• The amount of subcooling and vapor
bypass rates can only be determined
empirically. Simplified sizing procedures
are available [6, 30], but these are based
on heating all of the subcooled liquid to
its bubble point, and therefore are grossly
conservative. More reasonable sizing
criteria can be inferred from the principles
discussed in Ref. 9
Flooded drum scheme
Figure 7 illustrates the flooded reflux-drum
method. Here, the drum runs full of liquid,
and the level control of the reflux drum is
eliminated. Sometimes, especially in grav-
ity systems that have no reflux pump, the
reflux drum itself can be omitted. The pres-
sure controller directly controls distillate
flow. Due to the absence of vapor space,
the flooded drum is smaller than a drum
with a vapor space, the piping is simpler,
and together with the elimination of the level
control, this method can offer significant
capital-cost savings.
Due to the tight pressure control that is
usually required, distillate flowrate controlled
by this method is likely to fluctuate. These
fluctuations may destabilize downstream
units. This method should therefore only
be used when the product goes to storage
[1, 5, 20, 24], and should be avoided when
the product goes directly to another unit.
This method has sometimes been used
to control reflux flow, but this practice is not
recommended [5]. Here, reflux flow, rather
than product flow, is likely to fluctuate, and
this can destabilize the tower.
Besides the condenser venting issue
above, the flooded-drum method has an ad-
ditional venting issue. Accumulation of non-
condensables in the drum may unflood the
drum and interrupt the control action. These
non-condensables must be vented from near
the top of a liquid-full drum, so they need to
be vented to a facility that can handle liquids,
such as a knock-out drum. If accumulation
of non-condensables is infrequent, manual
venting from the top of the drum is often
sufficient to maintain satisfactory operation.
If non-condensables accumulate frequently,
or the column is run unattended, automatic
venting is required.
Figure 8 illustrates an automatic vent sys-
FIGURE 7. The flooded-drum
Over-
PC control method can offer cost
head savings in some applications,
from Vent but also introduces some
tower unique issues
Vent
Flooded
reflux drum
FC
Product
to
FC storage
Reflux to
tower
tem that has worked well in practice [5, 12,
31]. A second pressure controller (PC 2),
a level controller and a control valve in the
vent line are added. The setpoint of PC 2 is
lower than that of the normal pressure con-
troller (PC 1). When the drum is full, the level
controller keeps PC 2 tripped off, and the
vent valve is closed. Drum unflooding (due to
non-condensable accumulation) is sensed
by a drop in drum level. The lower level ac-
tivates PC 2. Since the setpoint of PC 2 is
lower than PC 1, it opens the vent valve. As
the drum pressure falls, PC 1 closes, help-
ing to build up the drum level. As soon as
the drum refills, the level controller trips PC
2, and the vent valve closes.
Since the venting required above is from
a liquid-full drum, leaks from the vent line
are likely to be liquid, giving a product loss
much greater than a leaking vapor valve.
Flashing of liquid across the valve can chill
the vent line, sometimes resulting in icing or
“weeping” on the outside of the pipe due to
atmospheric moisture condensation, with
possible overchilling or corrosion of the vent
FIGURE 8. This flooded-drum
pipe. Ref. 16 describes one case of icing automatic-vent system with
due to such a leak from a flooded drum in a dual pressure control has op-
debutanizer unit. erated well in practice [31]
~
~ Vent to
fuel
On/off
PC
2
Still PC
1
LC Flooded
reflux drum
Product
Reprinted courtesy
of AIChE

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2016 47

 PC
Over-
head
from
tower
Relief valve elevated
above condenser
liquid level
Vapor pocket
may form here
Vent
Flooded
reflux drum
FC
Product
to
FC storage
Reflux
to
tower
FIGURE 9. This configura- Unless the product is subcooled and at a
tion illustrates a relief valve significantly higher pressure than the storage
mounted on a flooded drum
facility, it is best to take the product to stor-
age from downstream of the reflux pump (as
shown in Figure 7). If the product is taken
directly from the drum, flashing may occur
downstream of the control valve, or it may
be difficult to get the product into storage
when the storage pressure is high. Either
may cause instability or back excessive liq-
uid into the condenser, thereby reducing its
capacity and possibly leading to a relief situ-
ation. In one depropanizer process, pressure
variations in elevated propane storage bul-
lets downstream induced intermittent flash-
ing and slug flow in the product line even
though the product was pumped [19]. Col-
lapse of vapor due to elevation and pressure
changes is believed to have caused transient
shockwaves and hammering, as well as
chattering of the relief valves, in the pump-
discharge circuit. The chattering was elimi-
nated by adding a backpressure controller
that prevented the flashing.
With the flooded drum method, a failure of
the reflux pump often produces a relief situ-
ation. The condensate has nowhere to go,
and quickly floods the condenser, ceasing
condensation, causing the tower pressure to
rise until the relief valve lifts. In other flooded
condenser schemes (for instance, Figure 2a)
where the reflux drum is not flooded, the
vapor space in the drum provides operators
48 CHEMICAL ENGINEERING
with a few minutes to take action before the
liquid fills the condenser. The author is familiar
with one plant that replaced a flooded drum
with a new non-flooded drum that was over
twice the volume (to accommodate a vapor
space) just to prevent recurrence of such a
relief scenario.
Any relief valve mounted on the flooded
drum (Figure 9) is most likely to discharge
liquid, which may not provide adequate
relief and may cause problems in the flare
system. Furthermore, when the drum relief
valve is elevated above the liquid level in the
condenser, a vapor pocket may form in the
valve inlet line during warm weather. Upon
cooling (for instance, at night), the vapor
pressure of the liquid in the small pocket
falls. The vapor pocket may collapse, forc-
ing a liquid rush that will hit the relief valve
and cause chattering, as reported in two
cases in Ref. 19. The need for a relief valve
on the drum should be critically reviewed,
as the relief valve on the tower should usu-
ally be able to relieve the drum.
Final remarks
Despite their importance in tower pressure
control, flooded condenser controls have
been some of the most challenging distilla-
tion-control techniques. However, operat-
ing experience indicates that they can be
quite trouble-free when correctly designed
and applied. It is hoped that the principles
and experiences described in this article
will pave the way for flooded-condenser
controls’ successful application for tower
pressure control. ■
Edited by Mary Page Bailey
References
1. Chin, T.G., Guide to Distillation Pressure Control Methods,
Hydrocarbon Processing, 58(10), p. 145, 1979.
2. Kister, H.Z., What Caused Tower Malfunctions in the Last 50 Years?,
Trans. Inst. Chem. Eng., Vol. 81, Part A, p. 5, January 2003.
3. Whistler, A.M., Locate Condensers at Ground Level, Pet. Refiner,
33(3), p. 173, 1954.
4. Schneider, D.F., and Hoover, M.C., Practical Process Hydraulic Con-
siderations, Hydrocarbon Processing, p. 47, August 1999.
5. Kister, H.Z., “Distillation Operation,” McGraw-Hill, New York,
N.Y., 1990.
6. Hollander, L., Pressure Control of Light-Ends Fractionators, ISA J.
4(5), p. 185,1957.
7. Sloley, A.W., Effectively Control Column Pressure, Chem. Eng. Prog.,
p. 38, January 2001.
8. Smith, C.L., “Distillation Control – An Engineering Perspective,”
Wiley, Hoboken, N.J., 2012.
9. Kister, H.Z., and Hanson, D.W., Control Column Pressure via Hot-
Vapor Bypass, Chem. Eng. Prog., p. 35, February 2015.
10. Kister, H.Z., and Hanson, D. W., Troubleshooting a Hot Vapor Bypass
Control Instability, 10th International Conference on Distillation and
Absorption, Friedrichshafen, Germany, September 14–17, 2014.
11. Kister, H.Z., and Litchfield, J.F., Distillation: Diagnosing Instabilities in
WWW.CHEMENGONLINE.COM JANUARY 2016
 the Column Overhead, Chem. Eng., p. 56, September 2004.
12. Kister, H.Z. “Distillation Troubleshooting,” Wiley – Interscience,
Hoboken, N.J., 2006.
13. Buckley, P.S., Luyben, W.L., and Shunta, J.P., “Design of Distillation
Column Control Systems,” Instrument Society of America, Research
Triangle Park, N.C., 1985.
14. Hollowell, R.W., Private communication, May 2015.
15. Lieberman, N.P., “Troubleshooting Process Plant Controls”, Wiley,
Hoboken, N.J., 2009.
16. Lieberman, N.P., “Process Equipment Malfunctions”, McGraw-Hill,
New York, N.Y., 2011.
17. Cardoso, R., and Kister, H.Z., Refinery Tower Inspections: Discov-
ering Problems and Preventing Malfunctions, Proceedings of the
Distillation Topical Conference presentation, AIChE Spring Meeting,
San Antonio, Tex., April, 2013.
18. Parsons, K., Private communication, March, 2015.
19. Mani, G., Avoid Condensation-Induced Transient Pressure Waves,
Hydrocarbon Processing, p. 41, January 2011.
20. Shinskey, F. G., “Distillation Control for Productivity and Energy Con-
servation”, 2d ed., McGraw-Hill, New York, N.Y., 1984.
21. Steinmeyer, D.E., and Mueller, A.C., Why Condensers Don’t Oper-
ate as They Are Supposed To, Panel Discussion, Chem. Eng. Prog.
70(7), p. 78, 1974.
22. Friedman, Y.Z., Condenser Hot Vapor Bypass Control, Hydrocarbon
Processing, p.150, June 2008.
23. Duguid, I., Take this Safety Database to Heart, Case MS16, Chem.
Eng., p. 80, July 2001.
24. Nisenfeld, A.E., and R.C. Seemann, “Distillation Columns”, Instru-
ment Society of America, Research Triangle Park, N.C., 1981.
25. Lupfer, D.E., Distillation Control for Utility Economy, Proceedings of
the 53rd Annual Convention, Gas Processors Association, March
25–27, Denver, Colo., p. 159, 1974.
26. Laird, D., and Cornelisen, J., Control-System Improvements Expand
Refinery Processes, Oil & Gas J., p. 71, September 25, 2000.
27. Sloley, A.W., Simple Methods Solve Exchanger Problems, Oil & Gas
J., p. 85, April 20, 1998.
28. Lieberman, N.P., “Troubleshooting Process Operations,” 4th Ed.,
PennWell Books, Tulsa, Okla., 2009
29. Hartman, E.L., and Barletta, T., Reboiler and Condenser Operating
Problems, PTQ, p. 47, Summer 2003.
30. Durand, A.A., Sizing Hot Vapor Bypass Valves, Chem. Eng., August
25, p. 111, 1980.
31. Kister, H.Z., and Hower, T.C., Unusual Case Histories of Gas Pro-
cessing and Olefins Plant Columns, Plant/Operations Prog.,6(3),
p. 151, 1987.
Author
Henry Z. Kister is a senior fellow and director
of fractionation technology at Fluor Corp. (3
Polaris Way, Aliso Viejo, CA 92625; Phone
+1-949-349-4679; Email: henry.kister@
fluor.com). He has over 30 years experience in
design, troubleshooting, revamping, field con-
sulting, control and startup of fractionation
processes and equipment. He is the author of
three books, the distillation equipment chap-
ter in Perry’s Handbook, and more than 100
articles, and has taught the IChemE-sponsored “Practical Distilla-
tion Technology” course more than 450 times. A recipient of several
awards, Kister obtained his B.E. and M.E. degrees from the Univer-
sity of New South Wales in Australia. He is a Fellow of IChemE and
AIChE, member of the NAE, and serves on the FRI Technical Advi-
sory and Design Practices Committees.

MONITOR VISCOSITY SIMPLY

SENSE MIXER MOTOR HORSEPOWER
WITH UNIVERSAL POWER CELL
EASY INSTALLATION PROFILING A PROCESS
•฀No฀holes฀in฀tanks฀or฀pipes฀ •฀฀Power฀changes฀refl฀ect฀viscosity฀changes
•฀Away฀from฀sensitive฀processes •฀฀Good฀batches฀will฀fi฀t฀the฀normal฀“profi฀le”฀for฀
that฀product
24
VERSATILE
POWER DECREASE
•฀฀One฀size฀adjusts฀to฀motors,฀from฀ 22
SHOWS BATCH
small฀up฀to฀150hp 20
IS DONE

•฀฀Works฀on฀3฀phase,฀fi฀xed฀or฀variable฀ 18

frequency,฀DC฀and฀single฀phase฀power 16
POWER
SENSOR
14
SENSITIVE
•฀฀10฀times฀more฀sensitive฀than฀ 12

just฀sensing฀amps 10 DRY MIX ADD LIQUID

HIGH SPEED LOW SPEED MIXER
CONVENIENT OUTPUTS 8 MOTOR

•฀฀For฀meters,฀controllers,฀computers฀ 6 BEGIN HIGH

SPEED MIX
4-20฀milliamps฀0-10฀volts 4

0
BATCH 1 BATCH 2 BATCH 3

CALL NOW FOR YOUR FREE 30-DAY TRIAL 888-600-3247

WWW.LOADCONTROLS.COM

Circle 7 on p. 66 or go to adlinks.chemengonline.com/61491-07

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM JANUARY 2016 49


Solve online analyzer time delays
by improving sampling system design
Process design engineers may occasionally become involved
in the design of online analyzers and their associated sampling
systems. Here are proven and practical strategies to signifi-
cantly reduce time delays without added or expensive invest-
ment. Time delay is key in sampling loop design: to use analyzer
data to adjust process control, it should reflect the most recent
process conditions. Specialist analyzer engineers are good at
specifying components in analyzer systems based on experi-
ence with previous analyzer systems. However, if they have
limited process engineering fundamental knowledge, analyzer
sampling loop design is often copied “as is” from prior designs
without understanding that a seemingly reasonable design may
cause significant time delays. Too often, the time delay is not
calculated, but is blindly assumed as acceptable by the process
engineers who design the main process system, causing unex-
pected process control problems in service.
WHAT ENGINEERING STANDARDS DO
AND DO NOT TELL YOU
Process design engineers who specify online analyzer sys-
tems must define the maximum time delay that the process
control can tolerate. The time delay should be designed as short
enough for process operators to promptly use analyzer data to
adjust process control, whenever analyzer readings are outside
of a normal range.
Typical engineering standards require less than a 60-second
time delay between the sample tap and the analyzer inlet. The
distance of the analyzer house from the sample tap is critical in
meeting this time constraint. Use small sample lines and small
volumes to minimize the residence time between the sample tap
and the analyzer.
Note that engineering standards may not always provide de-
tails about process engineering strategies to minimize time de-
lays. For example, they may state only the use of short distances,
small sample lines or fast loop to solve all time delay problems.
Standards also may not state that time delays occur not only
due to a time delay between the sample tap and analyzer inlet,
but also due to other factors, such as:
• Location of the sample tap relative to the point of
process control
Process Engineering
and Optimization
W. TANTHAPANICHAKOON and K. SURIYE,
SCG Chemicals, Thailand
• Type and size of the sample tap nozzle chosen
• Sample conditioning (e.g., filtering, coalescing,
vaporizing)
• Flow path components from the sample tap to analyzer
(e.g., flowmeter, pressure gauge)
• Analyzer response time and analysis time
• Sample switching for multiple samples using the same
online analyzer
This article provides proven and practical strategies to mini-
mize time delays with low or no investment.
Well-designed sample tap location, and tap nozzle type, can
significantly reduce time delay. For example, if a liquid sample
contains unwanted small particulates generated from a process,
then it is recommended to sample from an upward-flow verti-
cal process line to ensure a liquid-full condition inside the pro-
cess line. A 45° angle-cut probe, instead of a square-cut probe,
prevents unwanted particulates moving into the analyzer sam-
pling system. The right probe, properly installed, eliminates
the need to install a particulate filter system—a cause of time
delay and dead legs—in the analyzer sampling system.
Components on the flow path from the sample tap to the
analyzer, such as flowmeter, pressure gauge and temperature
gauge—used to monitor sample flowrate, pressure and tem-
perature—are also crucial parameters increasing time delay. It
is technically smart to move these components from the main
flow path to the sample bypass line or the process return line to
minimize time delay and dead legs.
Analyzer analysis time can also be a main cause of time de-
lay, especially on samples with a wide range of composition
(e.g., light gas to heavy liquid sample). Often, gas and liquid
phases are pre-separated and then analyzed individually by two
parallel analyzers to reduce the total time spent analyzing two
phases sequentially.
Therefore, the process design engineer must see the overall
big picture of time delays when designing the analyzer sampling
loop system. Overall time delay of an analyzer system—includ-
ing process delay, sample transport delay, sample conditioning
delay and analyzer response time delay—should be consistent
with the measurement and control requirements of a particular
process. Limiting the “sample conditioning” time delay to no
more than 30 seconds is recommended.
Hydrocarbon Processing | JANUARY 2016 71
 Process Engineering and Optimization

Two modes of sample transport from the sample tap to the
analyzer are:
1. Single-line transport
2. Fast-loop transport.
FIG. 1 shows single-line transport vs. fast-loop transport and
how the fast loop can reduce time delay. The fast loop allows the
use of a higher flowrate and higher sampling line size up to the
analyzer inlet. There, a specified amount of the fast-loop flow is
split and sent to the analyzer, while most of the fast-loop flow
bypasses the analyzer and returns to the process.
CALCULATING TRANSPORT TIME DELAY
OF AN ANALYZER SAMPLING LOOP
Steps to calculate transport time delay are as follows:
1. Specify components needed in analyzer sampling system
2. Specify a target transport time delay, such as 60 seconds
3. Identify sample’s flowrate required for analyzer +
analyzer bypass, used as basis for sizing sampling line
up to analyzer inlet (for single-line transport, analyzer
bypass flow = 0)
4. Identify available pressure drop for the sampling loop
5. Specify sampling line sizes for high velocity without
exceeding available pressure drop, and calculate
sampling line volume.
6. Calculate transport time delay = (sampling line
volume) ÷ (analyzer flow + analyzer bypass flow)
Single-line system:
Field station
Sample tap
Fast-loop system:
Sample tap
Sample tap
Fast-loop filter
Process analyzer
FIG. 1. Single-line vs. fast-loop sample transport system.
Process analyzer
7. If calculated time delay < target time delay, the
calculation is finished; if not, then go back to Step 3.
A common mistake in calculating time delays is the failure to
account for the following:
1. Gas (compressible fluid) velocity increase along the
gas sample line due to lower gas density when sampling
line pressure decreases as a result of pressure drop. This
expansion results in higher pressure drop along the line
as gas flows. However, liquid (incompressible fluid)
density remains constant, despite a drop in pressure.
Therefore, liquid velocity stays constant, resulting in
constant pressure drop per length.
2. Mixing volume time delay is at least three times the
average residence time. The mixing volume refers to
components with low length-to-internal-diameter ratio
(e.g., L/D < 10) such as filters, pumps and droplet
coalescers. When inlet concentration suddenly changes
to a new value, it theoretically takes at least three times
the average residence time in that mixing volume
for the outlet concentration to approach 95% of the
required concentration change. Therefore, in a time
delay calculation, always multiply the mixing volume
residence time by three. Special designs of filters, pumps
and coalescers may reduce this time delay effect.
FIG. 2 shows an example of a high-pressure reactor pilot
plant drawing a liquid sample from the inlet and a gas sample
from the reactor effluent. The reactor operates at 30 barg at a
temperature of 280°C. Single-line configuration is used with
flow and temperature monitors for both liquid and gas samples.
Both the liquid inlet and gas effluent’s flow into the analyzer
sampling system can be controlled using needle valves. A buf-
fer tank in the analyzer sampling system should be installed to
minimize fluctuation of sample flow and composition. Sample
pressure is regulated from 30 barg down to 5 barg with a pres-
sure regulator before the sample switching system. It is also rec-
ommended to use cascading double-block valve and bleed for
the sample switching system to eliminate dead leg problems.
Limit sample flows to the analyzer at not higher than 400 ml/
min for liquid and 2,000 ml/min for gas (analyzer flowmeter
condition = 25°C, 1 bara). TABLE 1 shows the total time delay in
this example calculated at 133.8 seconds for the liquid sample
and 375.7 seconds for the gas sample.

PROCESS ENGINEERING STRATEGIES

Liquid PI
inlet
Vent Analyzer TO REDUCE TIME DELAYS
TI
Buffer
If the calculated time delay is too long, follow the strategies
FI below:
tank
1. Regulate gas sample pressure at the field station
Pilot right after the sample tap. Regulating gas pressure
reactor
30 barg PI at the field station after the sample tap with a pressure
280°C TI regulator allows the gas volume to expand (lower
FI Buffer gas density) and velocity to increase. It reduces time
tank delays, but at the expense of higher pressure drop at
the same mass flowrate. The velocity increases by the
Gas absolute pressure ratio defined by the gas law equation.
effluent
Single-line system Sample switching system Pressure regulation is generally not desirable in liquid
lines. Define flowrate at analyzer conditions: make a
FIG. 2. Non-optimized pilot plant analyzer sampling loop system design.
correction for temperature and pressure at transport
72 JANUARY 2016 | HydrocarbonProcessing.com
 Process Engineering and Optimization

line condition. The temperature in the transport line composition to vary if the stagnant composition
can be assumed to be equal to ambient temperature in dead legs contaminate the main sample stream.
unless insulated or heat-traced, because the heat loss to Dead leg effects are more dominant on low-flow lines.
ambient is significant for small lines. Solutions are to:
[Time delay at Pline] = [Time delay at Panalyzer] × • Minimize dead legs in the flow path to analyzer
(Pline in bara ÷ Panalyzer in bara) × • Move dead leg components (e.g., pressure gauges,
(Tanalyzer in °K ÷ Tline in °K) pressure relief valves) to lines not flowing to the
To reduce time delay, move the pressure regulator after the analyzer, such as the sample bypass line.
sample tap point according to the design strategy in FIG. 3. TABLE 5. Heating or heat-tracing the sampling line. Gas
2 shows the time delay reduced below the non-optimized case in volume expands (lower gas density) with increasing
FIG. 2 by 20% for the liquid sample and by 64% for the gas sample. temperature, so velocity increases by the absolute
temperature ratio defined in the gas equation. Heating
Example 1. A fast-loop system requires 50 ml/min (1 bara, lines can sometimes eliminate the undesirable mixing
25°C) to the gas analyzer, and the sampling line volume from volume of a liquid knockout pot.
the sample tap to the analyzer is 300 ml. 6. Avoid adsorption on the tube inner surface.
• If the gas sample is tapped and transported close Measuring H2S at < a 10 mol ppm level can be
to analyzer conditions (1 bara, 25°C), what should inaccurate due to adsorption of H2S on the tube surface,
the analyzer bypass flowrate be to obtain a 60-second resulting in less or no H2S content making it to the
transport time delay? analyzer. Heating the tube to prevent adsorption or
[Answer: Analyzer bypass flowrate = 250 ml/min, coating the tube’s inner surface may help.
so that time delay at 1 bara = (sampling line volume)/ 7. Use process pressure to drive samples instead of a
(analyzer flow + analyzer bypass flow) = (300 ml)]/ sample pump. Whenever possible, first consider the
(50 ml/min + 250 ml/min) = 60 sec] use of process pressure by relocating the sample tap to
• If the gas sample is tapped and transported at average
4 barg and kept heated at 60°C until the analyzer inlet,
before dropping pressure across a needle valve into the Liquid PI Vent Analyzer
analyzer, then what would the time delay be? inlet TI
[Answer: Time delay at 4 barg = 60 sec × [(4 + 1) bara/ FI Buffer
1 bara] × [(273 + 25)°K/(273 + 60) °K] = 268.5 sec] tank
2. Change from single line to fast loop, if regulating
pressure is insufficient. Pilot
reactor
3. Avoid mixing volumes in the loop. The time delay 30 barg PI
effect of a mixing volume is three-fold, so it should be 280°C TI
avoided whenever possible. FI Buffer
tank
4. Avoid dead legs in the sampling line to analyzer.
Many components cause dead legs, such as pressure
gauges, pressure relief valves, lab sampling points, tee Gas
effluent
or purging connections, calibration or multi-stream Single-line system Sample switching system
manifolds. Dead legs not only increase sampling system
volume, causing time delays, but also cause sample FIG. 3. First optimized pilot plant analyzer sampling loop system design.

TABLE 1. Time delay of the non-optimized sampling system

Main components on the flow path that contribute to time delay
Non-optimized sampling system: Single-line system Sample switching
Sample Components 20 m of ⁄4-in. tube
1
Flowmeter—mixing volume Buffer tank**—mixing volume 6 m of 1⁄8-in. tube Totals
Condition 30 barg, 30°C 30 barg, 30°C 5 barg, 30°C 5 barg, 30°C –

Liquid inlet, Actual volume 240 ml 16 ml 196 ml 15 ml 467 ml

400 ml/min Corrected volume* 240 ml 48 ml 588 ml 15 ml 891 ml
Time delay 36.0 sec 7.2 sec 88.4 sec 2.2 sec 133.8 sec
Condition 30 barg, 30°C 30 barg, 30°C 5 barg, 30°C 5 barg, 30°C –

Gas effluent, Actual volume 240 ml 18 ml 196 ml 15 ml 469 ml

2,000 ml/min Corrected volume* 7,323 ml 1,619 ml 3,477 ml 86 ml 12,505 ml
Time delay 220.3 sec 48.5 sec 104.3 sec 2.6 sec 375.7 sec
*Notes: Details of corrected volume used in calculating time delay (1 and 2):
1. Correction for mixing volume = 3 x actual volume for the flowmeter and the buffer tank
2. Correction for pressure and temperature = actual volume × (Pline in bara/Panalyzer in bara) × (Tanalyzer in °K/Tline in °K) only for gas (Panalyzer = 1 bara, Tanalyzer = 25°C)
** Located after regulator

Hydrocarbon Processing | JANUARY 2016 73

 Process Engineering and Optimization

a higher pressure point and the analyzer effluent return
to a lower pressure location. Sample pumps tend to be
unreliable and create undesirable mixing volume. If a
pump is unavoidable, use a rugged high-quality standard
fast loop pump. For example, specify gear pumps or other
positive displacement pumps for liquid samples.
VELOCITY LIMITATION VS. PRESSURE
DROP LIMITATION
Two extreme cases of analyzer sampling loop design are:
1. Extremely long distance from sample tap to analyzer
2. Extremely small available pressure drop.
With extremely long distances, consider using a fast loop to al-
low higher velocity with a larger sampling line size without exceed-
ing available pressure drop. If the available amount of the sample is
too small for fast-loop configuration, choose to inject an inert dilu-
ent, such as nitrogen, to accelerate the sample transport right after
the sample tap through the long line to the analyzer. Note: It is
necessary to re-calibrate the analyzer for a nitrogen dilution case.
With an extremely small available pressure drop, avoid high-
pressure-drop needle valves and send the analyzer sample effluent
to a low-pressure location, such as a flare or vent to atmosphere—
if it is safe to do so. The process design engineer may also install
a sample pump. However, simply relocating the sample tap and
sample return locations may eliminate the low-availability-pres-
sure-drop problem. It is also important to prevent two-phase flow
formation—flashing liquid or condensing vapor—that leads to
excessive pressure drop or undesirable flow regime inside the
sampling line. Liquid slugs can also damage the analyzer.
BEWARE OF PHASE CHANGE
Sampling loop design is much more complicated when dealing
with sampling line phase change. Some phase changes are inten-
tional, while some occur by design mistakes. Three major types
of potential phase change that engineers should be aware of are:
1. Gas sample close to or at the dewpoint. Reducing gas
pressure right after the sample tap can make the gas move
away from the dewpoint and reduce the possibility of
condensation. Condensation can significantly change

TABLE 2. Time delay of the relocated pressure regulator case

Main components on the flow path that contribute to time delay
First optimized sampling system: Single-line system Sample switching
Flowmeter— Buffer tank—
Sample Components 1 m of 1⁄4-in. tube 19 m** of 1⁄8-in. tube mixing volume mixing volume 6 m of 1⁄8-in. tube Totals
Condition 30 barg, 30°C 5 barg, 30°C 5 barg, 30°C 5 barg, 30°C 5 barg, 30°C –

Liquid inlet, Actual volume 12 ml 46 ml 16 ml 196 ml 15 ml 467 ml

400 ml/min Corrected volume* 12 ml 46 ml 48 ml 588 ml 15 ml 891 ml
Time delay 1
1.8 sec 6.9 sec 7.2 sec 88.4 sec 2.2 sec 106.5 sec
Condition 30 barg, 30°C 5 barg, 30°C 5 barg, 30°C 5 barg, 30°C 5 barg, 30°C –

Gas effluent, Actual volume 12 ml 46 ml 18 ml 196 ml 15 ml 469 ml

2,000 ml/min Corrected volume* 317 ml 270 ml 313 ml 3,477 ml 86 ml 4,463 ml
Time delay1 9.5 sec 9.4 sec 104.3 sec 2.6 sec 133.9 sec
* Notes: Details of corrected volume used in calculating time delay
1
20% reduction in time delay for liquid inlet sample
2
64% reduction in time delay for gas effluent sample
** Located after regulator

TABLE 3. Time delay of the fast loop and new sample switching system
Main components on the flow path that contribute to time delay
Final optimized sampling system: Fast-loop system Sample switching
Sample Components 20 m of ⁄4-in. tube
1
5 m of 1⁄8-in. tube Totals
Condition 30 barg, 30°C at 600-ml/min flow 5 barg, 30°C at 400-ml/min flow –
Actual volume 240 ml 12 ml 252 ml
Liquid inlet
Corrected volume* 240 ml 12 ml 252 ml
Time delay1 24.0 sec 1.8 sec 25.8 sec
Condition 30 barg, 30°C at 8,000-ml/min flow 5 barg, 30°C at 2,000-ml/min flow –
Actual volume 240 ml 12 ml 469 ml
Gas effluent
Corrected volume* 7,323 ml 71 ml 7,394 ml
Time delay1 54.9 sec 2.1 sec 57.0 sec
* See notes in Tables 1 and 2
1
80% reduction in time delay for liquid inlet sample
2
85% reduction in time delay for gas effluent sample
** Located after regulator

74 JANUARY 2016 | HydrocarbonProcessing.com

the multi-component gas mixture composition to
the analyzer because heavy molecules are selectively
condensed more than light molecules, changing
gas composition.
2. Liquid sample close to or at the boiling point or
bubble point. Sub-cool the liquid before tapping the
sample to get below the bubble point and reduce the
possibility of flashing. Flashing alters liquid sample
composition because of selective loss of light molecules
into the vapor phase.
3. Liquid sample vaporization before entering the gas
analyzer. The sample vaporizer is a major cause of time
delays because vapor volumetric flow at the analyzer is
typically > 300 times the liquid volumetric flow at the
vaporizer inlet. Therefore, significant time delay occurs
on the liquid side upstream of the sample vaporizer.
Without proper measures, liquid-side time delay can
exceed 300 seconds. Recommended solutions are:
• Use small lines on the liquid side to minimize volume
and liquid residence time. The recommended practice
is a maximum 300-mm length of 1⁄8-in. tubing.
• Use the liquid bypass line to allow high liquid flowrate
up to the vaporizer inlet by splitting only small liquid
flow to the vaporizer and most of the flow to the
liquid bypass line.
• For clean and narrow-boiling-range liquid samples,
vaporize the liquid sample at the sample tap.
HPI MARKET DATA
2016 WEBCAST
>>View on Demand Today!
Discover what’s in store for the global refining, petrochemical and natural gas/LNG industries
in 2016 and beyond. View on demand at HydrocarbonProcessing.com/Webcasts
Lee Nichols
Editor/Associate Publisher,
Hydrocarbon Processing
“Global Construction and
Investment and Updates on
the Global Refining Industry”
Adrienne Blume
Executive Editor,
Hydrocarbon Processing
and Editor, Gas Processing
“Natural Gas/LNG Market
Status and Forecast”
Process Engineering and Optimization
Other practices for the sample vaporizer are to:
• Keep liquid pressure as high as possible at the
vaporizer inlet to avoid pre-flashing.
• Vaporize across a heated valve at the lowest pressure
and temperature possible for a given composition.
• Use electric heating with controlled temperature
as low as possible.
• Do not use steam heat due to the difficulty of
temperature control.
• Do not overheat the vaporizer body because the
inlet tube may get hot and flash some liquid before
entering the vaporizer, causing sample phase
separation and unexpected results.
CONCERNS OF SAMPLE SWITCHING DESIGN
Typically, one analyzer is designed to analyze multiple sam-
ples from different sample taps, so sample switching is often
required. Three conventional sample switching configurations
(FIG. 4) are:
1. Conventional double block and bleed
2. Cascading double block and bleed
3. Outlet flow loop double block and bleed.
Advantages of conventional double block and bleed are cost
and ease of installment. A disadvantage is the presence of dead
legs, so cascading double block and bleed is preferred, even
with a higher installation cost. However, a much longer time
delay becomes the next problem for cascading double block
Ben DuBose
Digital Editor,
Hydrocarbon Processing
“Petrochemical Outlook
for 2016 and Beyond”
Mike Rhodes
Managing Editor,
Hydrocarbon Processing
Hydrocarbon Processing | JANUARY 2016 75
 Process Engineering and Optimization

1) Conventional double block and bleed 2) Cascading double block and bleed 3) Outlet flow loop double block and bleed
Vent Vent Vent
Sample 1

Sample 1 Analyzer Analyzer Sample 1

Sample 2

Sample 2 Sample 2 Analyzer

Sample 3

Sample 3 Sample 3

FIG. 4. Sample switching configurations.

MANAGING TIME DELAYS IN A NUTSHELL

Vent
Liquid PI
inlet
TI PI
FI
Pilot
reactor PI
30 barg
280°C
TI PI
FI
Gas
effluent
Fast-loop system Sample switching system
FIG. 5. Final optimized pilot plant analyzer sampling loop system design.
and bleed. To overcome these limitations, it is recommended to
use an outlet flow-loop double-block-and-bleed configuration
to significantly reduce both time delay and dead leg problems.
FIG. 5 shows the final optimized analyzer sampling loop sys-
tem with various strategies implemented. A fast-loop configura-
tion is used with the removal of the original buffer tank. An out-
let flow-loop double-block-and-bleed configuration reduced
dead legs and time delay. It also helped minimize the fluctuation
of the sample’s flow and composition after the removal of the
buffer tank. The pressure regulator was moved to the position
right before the sample switching system to not limit fast-loop
high flowrate. Components were also moved from the main
flow path to the fast-loop return line. There were limited sample
flows to the analyzer at not higher than 400 ml/min for liquid
and 2,000 ml/min for gas. TABLE 3 shows that the time delay
was reduced below the non-optimized case in FIG. 2 by 80% for
the liquid sample and 85% for the gas sample, in which the final
time delays for both the liquid inlet and gas effluent are within
the typical engineering standards at 60 sec.
Analyzer
Time delays can be managed by using process engineering
fundamentals to analyze the problem and correct root-cause
phenomena, not only the problem symptoms.
Process design engineers should never blindly believe that
analyzer engineers or other discipline engineers are doing ev-
erything right. It is the duty of competent process design engi-
neers to cross-check the design from other discipline engineers
by studying relevant engineering standards and utilizing sound
process engineering fundamental knowledge.
REFERENCES
1
T. Waters, Industrial Sampling Systems, Swagelok, 2013.
2
Instrument Engineers’ Handbook, 4th Edition, Chapter 8, ISA, 2003.
3
Pichler, H. and O. Sauer, “Consider investing in a standard compliant process
analyzers,” Hydrocarbon Processing, July 2009.
4
Benton, A. and C. Valiz, “Use online analyzers for successful process monitoring,”
Hydrocarbon Processing, January 2011.
WIROON TANTHAPANICHAKOON is a process technology
leader for the Process Technology Center of SCG Chemicals Co.
Ltd. in Thailand. He has authored several international papers
and technical articles and holds technology patents. He is
now in charge of developing and implementing in-house and
external process technologies for SCG Chemicals. His expertise
also includes process and equipment design plus the
conceptual, basic and detailed design of new processes. He has had experience
as a process engineer with an ExxonMobil subsidiary refinery in Thailand and as
an energy improvement engineer for SCG Chemicals. He obtained his BE and
ME degrees in chemical engineering from Kyoto University, Japan, and is a senior
licensed engineer in Thailand and a senior member of AIChE. He was also a
member of the Elsevier Editorial Board in Process Plant Design during 2014–2015.
KONGKIAT SURIYE is a senior researcher for the Research and
Development Center of SCG Chemicals, Co. Ltd. in Thailand.
He is now a group leader for the advanced catalytic process
research team. He has authored more than 30 international
papers and one book chapter in the fields of heterogeneous
catalyst and catalytic processes, and has invented several
patents in these fields. One of his developed technologies has
been implemented on a commercial scale, and three are now at pilot scale. His
expertise also includes laboratory and pilot-scale reactor design and scale-up,
and also a design of reaction online analysis with sampling loop systems. He
obtained his BE degree (1st ranking in class) in chemical engineering from King
Mongkut’s University of Technology in Thonburi, Thailand, and also a PhD degree
in chemical engineering from a joint program between Chulalongkorn University
in Thailand and the University of California–Davis, in the US.

76 JANUARY 2016 | HydrocarbonProcessing.com
 Cover Story

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

A BRIEF HISTORY OF design. However, many have 0.7

PROCESS CONTROL never considered or studied
0.6
DESIGN BASIS AND
process dynamics. Process
engineers often provide the 0.5
PROCESS VARIABILITY
preliminary instrumentation and 0.4
CONTROL BASICS control requirements for new 0.3
ADVANCED PROCESS projects. Control engineering is
0.2
CONTROL just the next step. Control engi-
neers try to identify and under- 0.1
BATCH PROCESS
CONTROL
stand sources of process vari- 0
0 20 40 60 80 100 120 140 160 180 200 220 240
ability that can impact product Time, s
PROCESS SAFETY IN quality, and then reduce the FIGURE 1. Responses to process inputs in self-regulating processes can take the
CONTROL variability to mitigate its adverse form of first-order plus deadtime (FODT) or second-order plus deadtime (SODT)
PROCESS DATA economic effects.
There are many renowned chemical engi- Electronics gave the advantage of faster
CONCLUDING REMARKS
neers who have made careers and reputa- communications between field instruments
tions for themselves as control engineers, in- and controllers located in a central control
cluding the prolific author Greg Shinskey, the room, as well as space savings and some
father of model-predictive advanced control, improved features. They generally mimicked
Charles Cutler, and academics like Thomas pneumatic controls, but electrical engineers
Edgar, Thomas McAvoy and Dale Seborg. began to displace mechanical engineers as
Even if a process engineer never becomes control engineers. Relay systems were used
a control engineer, being aware of process to provide interlocks, logic control and se-
control requirements and challenges will lead quence control.
to better process designs. This article pro- Inevitably, analog single-loop controllers
vides information to aid chemical engineers in were replaced with multi-loop digital control
their understanding of how to reduce process systems — first, control computers and later,
variability by better controlling processes. distributed control systems (DCS). Also at this
time, programmable logic controllers (PLCs)
A brief history of process control began to displace systems of relays for logic
Early process controllers were mechani- and sequential control. Modern control sys-
cal devices using pneumatics and hydrau- tems are now leveraging the Internet, wireless
lics. Mechanical engineers were common technologies and bus technologies in new
in control engineering, especially since the and effective ways. Field instruments and
most common final control element — the final control devices are becoming increas-
control valve — is inherently a mechanical ingly more “intelligent,” providing for more
device. Pneumatic controllers were gradu- non-control information than control informa-
ally replaced with analog electronic systems. tion. Tools and user interfaces are becoming
42 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017
friendlier to use and more capable, offer-
ing tremendous productivity gains.
Electrical engineers continued to
dominate the field of automation and
control. But the additional computing
capability in microprocessor- and com-
puter-based systems led to an oppor-
tunity for more advanced control strat-
egies. Chemical engineers, with their
knowledge of process behavior and
process requirements, have been work-
ing for years in the field of automation
and control, and are often able to gain
more value out of control systems than
others with less process understanding
might achieve.
Design basis and variability
Process engineering focuses on process
design, and defines or assumes a de-
sign basis. That basis typically includes
normal, maximum and minimum pro-
duction rates, and the process engineer
tries to optimize the process design, first
in terms of capital cost and second in
terms of operability. At this stage, project
cost considerations and the availability
of standard process equipment may re-
quire design compromises that lead to a
process design with control challenges.
The design basis is a guideline, but
operating conditions in a commissioned
plant may change over time. Equip-
ment (especially control valves) wear,
feedstock qualities vary, catalysts age,
processes are impacted by varying am-
bient conditions, and other sources of
variability impact production. Market
and regulatory conditions may also
vary, shifting demand for certain prod-
ucts and byproducts or penalizing the
production of waste products. The con-
trol system of the plant is intended to
mitigate the effect of incoming sources
of variability on product quality variabil-
ity. As plants become increasingly com-
plex, operators are faced with bigger
challenges, and simply operating the
process manually is no longer an op-
tion. A frequently cited analogy is the
pilot in an advanced meta-stable jet de-
pendent on advanced avionics.
Perhaps the best way to look at au-
tomation and control is as the business
of managing process variability in real-
time. One important thing to understand
is that control systems are generally able
to attenuate low-frequency (“slow”) vari-
ability (on the order of seconds, minutes
or more), but cannot attenuate high-
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
FODT vs INPUT
(Non-integrating)
1.0
0.9
Percent change in process output
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120 140
Time, s
160 180 200 220 240
frequency (“fast”) variability (less than a FIGURE 2. Typical first-order plus deadtime
few seconds). (FODT) responses are characterized by a
rapid initial response to a process input,
Fortunately, process design can often followed by slowing response as a new
be used to attenuate fast variability. steady state is reached
Surge vessels can be used to attenuate
highly variable flows between units, for
example, reducing the disruption to the
downstream unit from variability in the
upstream unit.
Control engineers need to understand
process dynamics, a topic area that is
not always considered as part of the
core of process design. It is convenient
to think of process dynamics in terms
of process inputs and process observa-
tions. Process inputs are material or en-
ergy flows, and they may be flows into,
out of, or intermediate within a given
process. As flows are changed, the
process is affected, as seen by process
observations. Process observations are
measured as variables like tempera-
tures, pressures, levels, compositions
and flowrates.
As process designs are optimized
for energy recovery and minimization
of both capital cost and operating cost
for a plant, they incorporate increasing
integration between process streams.
If variability is not controlled in a highly
integrated process with a high degree
of process interactions, there are more
pathways for it to create quality issues.
Therefore, it is increasingly important for
design and control engineers to work
together to ensure operability and strat-
egize how to attenuate variability.
Control basics
Most process responses can be classi-
fied into self-regulating and non-self-reg-
ulating (or integrating). Self-regulating pro-
cesses respond to a change in process
MARCH 2017 43
 TUNING A PROPORTIONAL-INTEGRAL-DERIVATIVE (PID) LOOP

P
ID controllers are defined by the control algorithm, which gener-
ates an output based on the difference between setpoint and
process variable (PV). That difference is called the error, and the
most basic controller would be a proportional controller. The error is
multiplied by a proportional gain and that result is the new output. The
proportional gain may be an actual gain in terms of percent change of
output per percent change of error or in terms of proportional band.
Proportional band is the same as gain divided by 100, so the effect is
the same, even if the units and value are different. When tuning a control
system, it is important to know whether the proportional tuning param-
eter used in the controller being tuned is gain or proportional band.
When the error does not change, there is no change in output. This
results in an offset for any load beyond the original load for which the
controller was tuned. A home heating system might be set to control
the temperature at 68˚F. During a cold night, the output when the
error is zero might be 70%. But during a sunny afternoon that is not
as cold, the output would still be 70% at zero error. But since not
as much heating is required, the temperature would rise above 68˚F.
This results in a permanent off-set.
Integral action overcomes the off-set by calculating the integral of
error or persistence of the error. This action drives the controller error
to zero by continuing to adjust the controller output after the propor-
tional action is complete. (In reality, these two actions are working
in tandem.) The integral of the error is multiplied by a gain that is
actually in terms of time. Again, different controllers have defined the
integral parameter in different ways. One is directly in time and the
other is the inverse of time or repeats of the error per unit of time.
They are functionally equivalent, but when calculating tuning param-
eters, the correct units must be used. Adding further complication,
the time can be expressed in different units, although seconds or
minutes are usually the design choice.
And finally, there is a derivative term that considers the rate of
change of the error. It provides a “kick” to a process where the error
is changing quickly and has a gain that is almost always in terms of
time. However, again the units of time may be seconds or minutes.
Derivative is not often required, but can be helpful in processes that
can be modelled as multiple capacities or second order. Derivative
action is sensitive to noise in the error, which magnifies the rate of
change, even when the error isn’t really changing. For that reason,
derivative action is rarely used on noisy processes and if it is needed,
then filtering of the PV is recommended. Since a setpoint change
can look to the controller like an infinite rate of change and pro-
cesses usually change more slowly, many controllers have an op-
tion to disable derivative action on setpoint changes and instead of
multiplying the rate of change of the error, the rate of change of the
PV is multiplied by the derivative term.
There are two steps to tuning a controller. First the process dy-
namics must be identified. This can be done with an open-loop or
closed-loop step test. In open loop, the controller is put in manual
mode and the output is stepped. The PV is observed and the pro-
cess deadtime, gain, and time constants are estimated. Several
steps should be made to identify any nonlinearity and to ensure the
response is not being affected by an unmeasured disturbance. In
closed loop, the controller is forced to oscillate in a fixed cycle by
stepping the output, forcing it to oscillate with an amplitude that
will be dependent on the process gain and step size. This can be
achieved with a controller by zeroing the integral and derivative
terms and adjusting the proportional gain until the cycle is repeat-
ing, or by using logic that switches the output when the cycling PV
crosses the setpoint value.
The second step is calculating the tuning parameters. There are
different guidelines proposed by different authors and even software
that will calculate the tuning parameters for the tuner to achieve
the desired response. One guideline that is wisely favored is the
“lambda” tuning method. Lambda refers to the closed loop time
constant in a controller response. The advantage of this kind of tun-
ing is that the tuner is free to choose the speed of response or the
aggressiveness of the controller tuning. There is a tradeoff in loop
tuning. As noted earlier, faster response or more aggressive tuning
may result in some overshoot or even cycling response that is un-
desirable and the loop could become completely unstable if there is
any nonlinearity in the process. Therefore, robustness is the sacrifice
for more aggressive control and lambda can be used to strike an
optimal balance between robustness and aggressiveness.

input by settling into a new steady-state
value. For example, if steam is increased
to a heat exchanger, the material being
heated will rise to a new temperature. Re-
ponses often take the approximate form
of first order plus deadtime (FODT) or
FIGURE 3. Integrating, or non-self-regu-
lating process variables do not settle into
second order plus deadtime (SODT) (Fig-
a new steady-state value within allowable ures 1 and 2). In a heat exchanger, for ex-
operating limits ample, when the steam valve is opened,
SODT vs INPUT
(Integrating)
0.24
0.22
0.20
Percent change in process output
more steam enters the heat exchanger.
First, the steam pressure in the exchanger
rises and heat transfers to the tubes and
finally to the colder stream. The tempera-
ture of the cold stream takes some time
before it begins to rise. Then it rises grad-
ually and increases its rate of change until
it approaches the new steady-state tem-
perature, where the temperature rise be-
gins to slow. The characteristic response
is a SODT.
The steam flow began increasing as
soon as the valve started moving. But if
a controller was telling the valve to open,

0.18 there might have been a short delay be-

0.16 fore the valve actually moved and the
0.14 steam flow changed. The steam flow
0.12 begins to increase quickly and begins to
0.10 increase more slowly as the new steady-
0.08 state flow is achieved. This is a typical
0.06 FODT response.
0.04 Self-regulating control loops can be
0.02 tuned for closed-loop control response
0.00 to assure that the process observation
0 20 40 60 80 100 120 140 160 180 200 220 240
Time, s (sometimes known as the process vari-
44 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017
able, or PV) is driven to and main- mon, and when it occurs, it tends to
tained at its target setpoint. The con- be minimal.
trol response can be tuned for faster Fortunately, controllers can also
or slower response, but as the speed be tuned on integrating processes
of response increases, so does the to achieve a first-order response.
risk of overshoot or oscillation. However, the response does not
Different measures of performance look exactly like the response of a
have been developed with tuning self-regulating process. Following a
rules to approximately achieve these setpoint change, the PV will move
objectives. Early performance objec- to the new setpoint and overshoot
tives focused on minimizing error,
square of the error, or absolute error
over time. Tuning to achieve quar-
ter-amplitude damping was often
described in early control literature.
Zeigler-Nichols tuning rules were
proposed to achieve this kind of re-
sponse. But this kind of aggressive
tuning results in some cycling.
Recent thought in automation
GEMÜ Quality Products...
prefers to attenuate variability, and GEMÜ Quality Service
that includes closed-loop oscilla-
4212 & 4242 Automation features:
tion, so most loops today should be
• Class 1 DIV 2, UL & CSA
tuned for a first-order response, with
• Multiple control options: 24 VDC, AS-I, DeviceNet
the response time being defined ac-
• Super bright LEDs indicate valve position
cording to process requirements.
The fastest-responding loops are Multiport Diaphragm Valve features:
limited by the point of critical damp- • Thousands of block designs to solve process challenges
ing. However, where loops inter- • Minimize dead leg and hold up volume for
act, one can be slowed relative to optimal process efficiency
another to prevent the interacting
loops from fighting with each other. Globe Valve features:
• High cycle life 550
Important time constants are the Globe valve
deadtime, which is the time before • Variety of end connections, materials with 1434
of construction and actuation positioner
the process observation is observed
to change, and a first-order time
constant, which is the time it takes
once the process begins moving to 4212
achieve approximately 63% of the Switch
way to the target setpoint. A first-
order process normally takes about
four time constants, plus the dead-
time, to reach steady-state at the
target setpoint.
Non-self-regulating process
variables do not settle into a new
steady-state value, at least not
within allowable operating limits 650 Diaphragm
Multiport valve with 4242
(Figure 3). Changing the rate of feed
into a vessel will change the rate at
which the level rises (or falls). In the
absence of some kind of control, the See us at Interphex Booth 2663i
level would continue to rise until the
tank overflowed. Usually integrat- 3800 Camp Creek Parkway • Building 2600 • Suite 120
ing process responses can be de- Atlanta, GA 30331 • info@gemu.com • Tel: 678-553-3400
scribed as deadtime and integrating www.gemu.com
gain or ramp rate. Sometimes, there
may be a lead or lag associated with
the ramp rate, but this is not com- Circle 22 on p. 90 or go to adlinks.chemengonline.com/66426-22

FKH(QJB*(08BLVOQGLQGG 30
45
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017

Elements of a feedback control loop
Load
Set point Error Output Final control element or Fluid
∆ Controller 
slave loop process
Sensor/
Process value transmitter
FIGURE 4. Most process control is accom- slightly before turning around and set-
plished by using a feedback control loop tling back at the target value. Following
a disturbance, the PV will deviate from
the target setpoint until finally being ar-
rested and returning to setpoint. Dead-
time is the same as for a self-regulating
process. There is no open-loop time
constant by definition. However, the
closed-loop time constant for an inte-
grating process is defined as the time
it takes to first cross the target setpoint
following a setpoint change or the arrest
time for a disturbance. An integrating
process normally takes about six time
constants, plus the deadtime, to reach
steady-state at the target setpoint fol-
lowing either a setpoint change or a load
disturbance.
Feedback controllers
Process control usually takes the form
of a feedback controller (Figure 4). Some
process inputs can be manipulated in
order to drive important process obser-
vations to targets or setpoints. Other
process observations may not be con-
trolled to a target setpoint, but they are
not allowed to exceed upper or lower
constraint limits. A control-loop includes
a measurement of the process obser-
vation to be controlled (the PV), a final
control element (usually a control valve)
that varies the process flow to be ma-
nipulated and a controller that makes a
move based on where the process ob-
servation is relative to its setpoint.
The workhorse controller in the pro-
FIGURE 5. Several aspects of a process-
control loop can influence performance cess industry remains the PID (propor-
and profitability tional-integral-derivative) controller. It is
Feedback control: What in this loop limits profitability?
Operator practices? Dead band? Vessel design?
Control strategy? Sizing?
Flow characteristic?
Load
Set point
∆
Error Output Final control element or Fluid
Controller
slave loop  process
Sensor/
Process value transmitter
Tuning?
Algorithm?
Nonlinearity?
Sampling, filtering?
46 CHEMICAL ENGINEERING
robust and a good fit for the job as long
as the process response is not exces-
sively nonlinear or characterized by a
Product dominant deadtime dynamic. Propor-
tional, integral and derivative are the ac-
tions the controller can apply to drive the
PV to setpoint. Every controller manufac-
turer may employ a slightly different form,
structure and options, but the functional-
ity and results are the same. The propor-
tional, integral and derivative parameters
can be adjusted by the control engineer
to provide the best controller response.
In order to properly tune a control loop,
it is necessary to understand the things
that influence loop performance and
process profitability (Figure 5).
Often, process inputs can impact
more than one important process obser-
vation. If the heat exchanger was the re-
boiler of a distillation column, increasing
the steam could affect the levels in the
base of the column and the reflux accu-
mulator and compositions at the top and
bottom of the column. It might also af-
fect the column pressure and differential
pressure, and will affect temperatures up
and down the column. Similarly, a pro-
cess observation might be affected by
more than one process input. The dis-
tillate composition may be affected by
the steam flow to the reboiler, the reflux
flow, the feed flow, the product flows
and other process inputs. An interac-
tive process requires that the controls
be designed to minimize the detrimental
impact of multivariable interaction, where
two or more loops could fight with each
other. One way to do this is with a de-
coupling strategy, which is something
easily understood by process engineers.
Feed-forward control and sometimes
ratio-control strategies are used to de-
couple process interactions. The inter-
acting process inputs may be controlled
or could be “wild” disturbances. Another
way to decouple loop interactions is by
tuning one loop for a relatively faster
response and the other for a relatively
slower response. This technique is very
effective and is naturally applied when
tuning cascade loops.
Product Advanced process control
There are different definitions of ad-
vanced process control (APC). Some
people consider ratio control and over-
ride control to be advanced control.
Control strategies that use feed-for-
ward, override-control, cascade and
WWW.CHEMENGONLINE.COM MARCH 2017
ratio loops and other complexities are
often referred to as advanced regulatory
control (ARC). Another type of control-
ler is the multivariable, model-predictive
controller (MPC). The response models
of all process outputs to changes in any
process inputs are modelled and incor-
porated into the controller.
The controller attempts to maintain
the controlled variables at targets and
constraint variables within limits while
minimizing the moves of process inputs.
It does this with an algorithm that con-
trols a prediction sometime in the future
rather than the current process value. It
is an ideal approach for interactive prob-
lems, since instead of decoupling the
interactions, it coordinates the moves
to compensate for or accommodate the
known interactions. Those interactions
are identified in the embedded models.
It is also the only truly effective means of
handling deadtime-dominant processes
because the deadtime is inherently de-
fined in the controller models.
APC has been widely applied in the
petroleum refining industry and is gain-
ing greater acceptance in other in-
Eliminate Downtime & Minimize Maintenance Costs
Reduce Energy Consumption & Emissions
Maximize Productivity
Circle 34 on p. 90 or go to adlinks.chemengonline.com/66426-34
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
$$$
Limit or spec
Setpoint
Before APC After APC
dustries. It is an excellent platform for FIGURE 6. Advanced process control (APC)
constraint optimization. Many process- techniques help allow processes to oper-
ate closer to the limit through constraint
control problems benefit from constraint optimization
optimization. Optimization objectives,
such as maximizing production and
yield, and minimizing give-away and
energy consumption are examples of
where constraint optimization can gen-
erate substantial benefits over single-
loop control. Maximizing or minimizing
some variables can drive the process to
constraint limits and the models allow
for tight control at constraint limits with-
out violating them (Figure 6).
Batch process control
Up to this point, the discussion has cov-
ered continuous processing. Continu-
ous processes dominate the chemical
process industries (CPI), but some sec-
tors of the CPI, including pharmaceuti-
■ Universal corrosion resistance in extremely
high temperatures.
■ Custom complex shapes in any size with tight
tolerances.
Visit our NEW website: http://hexoloy.com
For more information, technical data & case studies contact:
SCD.sales@saint-gobain.com ■ 716-278-6233
MARCH 2017 47
 GLOSSARY OF CONTROL TERMINOLOGY
APC Advanced process control. A general term for types of con-
trols more elaborate than the basic loops. MPC is a type of
APC, and MPC is often used interchangeably with APC. Other
examples of APC include fuzzy logic and expert systems
ARC Advanced regulatory control. A complex control strategy
often involving more than one PID controller (examples in-
clude cascade control, ratio control, feed-forward control,
override control and inferential control)
CV Control variable. A process observation that has a setpoint
that may be provided by the operator or by a supervisory
controller
DCS Distributed control system. A digital process control plat-
form in which various controllers are distributed throughout
the system
DV Disturbance variable. Measured process inputs that are
not manipulated by the controller (also known as feed-for-
ward inputs)
HC Hand control. Used to indicate a manually positioned valve
LV Limit variable. A constraint variable or process observation
without a setpoint
MIMO Multi-input, multi-output. A term describing a multivariable
controller
MISO Multi-input, single output. A term describing a multivari-
able controller with one output
MPC (1) Model predictive controller. A controller that controls
future error rather than current error. To do this, it incorpo-
rates process response models that describe the dynamic
behavior of process observations (CVs and LVs) to changes
in process inputs (MVs and DVs). This allows the controller
to plan moves in the future to provide control over a future
time interval. This technology is helpful with processes that
have long process delays or a high degree of interaction be-
tween multiple process inputs and outputs. It is the ideal
platform for constraint optimization
(2) Multivariable process control. Means controlling more
than one measurement or variable at a time from the same
calculation. Since most model-predictive controllers are also
multivariable controllers, the definitions are often used inter-
changeably
MV Manipulated Variable. A process input that is an output of
a controller
PID Proportional, integral, derivative. The name used for the
most common control loop algorithm seen in the process
industries. PID controllers are SISO feedback controllers
PLC Programmable logic controller. Computer for industrial
control
PV Process value. The term commonly used for the CV of a
PID control loop
SIMO Single-input, multi-output. A term describing an uncom-
mon multivariable controller with one process output (an ex-
ample would be split-range control)
SISO Single-input, single-output controller. A term describing
typical PID controller or any controller with one input and one
output
SP Set point. The target value for the control variable
TSS Time to steady state. The time required for a self-regulating
process to come to a new stable steady state after a change to
a process input

cals and specialty chemicals, rely heavily
on batch processing. Some engineers
muse that all processes are batch pro-
cesses, but some batches are longer
than others. A batch control engineer
might suggest that a batch process is
just a continuous process that never
gets the chance to reach steady-state.
Both are valid points of view. Designing
batch process sequences and recipes
fall right in the comfort zone for chemi-
cal engineers. But the more interesting
part of batch control is not defining the
normal sequence of steps. Rather, it is
defining what should happen when an
abnormal event occurs. Can a batch be
“saved” following an upset or must it be
scrapped? What is required to rework a
batch that suffered an abnormal upset?
Thinking through the possible problems
that could disrupt a batch process and
defining safe sequences to abort or re-
cover a batch are classic chemical engi-
neering exercises.
Another opportunity for batch optimi-
zation involves trying to minimize transi-
tion times between steps. This can be
done with equipment selection, but also
with logic in the batch sequence. Ramp
rates and dwell times can be minimized
to the extent practical without impacting
48 CHEMICAL ENGINEERING
batch quality. There is a standard de-
fined (ANSI/ISA S88) for batch process
control that standardizes the concepts
of control, equipment and unit modules
in a batch process.
And batch processes are often de-
signed to make a variety of products or
product grades. Furthermore, there may
be multiple trains of equipment with some
common process equipment or utilities.
These plants may involve special recipes.
Recently in batch control, the focus has
been on managing multiple recipes and
optimizing equipment selection for maxi-
mum or optimum production.
Because product flaws in the phar-
maceutical industry can be devastating,
traceability is a major concern. This in-
cludes traceability of the materials con-
sumed in the production of pharmaceu-
ticals, and traceability of the equipment
and processes used to produce the
pharmaceuticals. Regulatory involve-
ment is high, and validation is an inte-
gral part of pharmaceutical processes.
This requires more data collection and
more rigorous adherence to manage-
ment of change (MOC) procedures than
most other processes, whether batch
or continuous.
WWW.CHEMENGONLINE.COM MARCH 2017
Process safety in control
Another area of process control deals
with safety instrumented systems (SIS).
Up to this point, the discussion has
centered on control requirements to
keep the process running in the face of
variability. Safety systems have a single
function, which is to safely shut down
a process if a catastrophe is imminent.
Process engineers may be better pre-
pared to consider process safety than
most disciplines, at least with regard to
the CPI. The general concept is to evalu-
ate the risks, in terms of probability, and
the magnitude of the consequences.
Layers of protection are defined and
deployed to reduce the risk of a serious
safety or environmental exposure. High-
risk possibilities need to employ engineer-
ing solutions to reduce the risk. Some
solutions will include process design,
such as dikes around tanks and pressure-
relief equipment. Controls will also be em-
ployed to reduce the risk, including safety
interlocks. The requirement for high on-
demand availability of the safety-protec-
tion systems leads to specialized safety
systems with redundancy (including triple
redundancy) and pro-active diagnostics
to monitor the health of the safety sys-
tems. One of the first layers of protec-
tion is alarm management, although it is
limited by the presence of the human el-
ement to respond to an alarm (for more
on alarm management, see Chem. Eng.
March 2016, pp. 50–60). Designing
safety-instrumented control systems is a
specialized area that is critical in managing
the risk of hazards in the process. There is
a growing trend to design safety systems
to be integrated into — but still separate
from — the basic control systems. Care
is taken during design to ensure the inte-
gration does not create a vulnerability or
common point of failure of the safety sys-
tem function and reliability.
Along with this trend is the increasing
use of diagnostics and capabilities of
“smart” instruments and field devices to
reduce the probability of failure on de-
mand. This is a critical consideration for
safety systems, because they may not
be employed for long periods of time,
if ever, but then must work when called
upon to shut a process down safely.
Process data
A final area of process control where pro-
cess engineers can have a significant im-
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
pact is data management and analysis.
Control systems have access to a great
deal of data other than control data. His-
torization and archiving of process data
enables process engineers to identify
and prioritize continuous improvement
opportunities and allows management
to make more effective decisions regard-
ing operation and future investment.
Business systems that manage main-
tenance processes, quality processes,
planning processes and other work pro-
cesses can be integrated with process
control. This has been enabled by modern
technology for networking, databases,
operator interfaces and enterprise-man-
agement software all working together.
While the nature of integration of these
various systems requires more knowledge
in computer programming, database ad-
ministration and networking than chemical
engineers might learn in their academic
programs, the process management re-
quirements require an understanding of
the process plant and its economic sen-
sitivities. Chemical engineers are likely to
have a better understanding than most
of the information required by company
managers at both the local and corpo-
rate level in order to make best use of the
data and systems in place. The increasing
wealth and richness of data makes analy-
sis of that data with evolving “big data”
tools a real opportunity. Networking, data
sharing, and collaboration between the
plant and specialized resources located
far away is the promise of the Industrial
Internet of Things (IIOT).
Concluding remarks
Often the greatest knowledge gap for
a chemical engineer who wants to be-
come a process automation engineer
is deep knowledge of instrument and
control hardware. This is not an insur-
mountable problem, however, because
vendors are happy to share the infor-
mation you need. A good salesperson,
perhaps contrary to popular opinion,
can be a valuable and trusted advisor.
The best salespeople know that exag-
gerating the benefits of one offering for
immediate sale may win the order, but
will lose the confidence and trust of the
customer for future opportunities.
In most cases, vendors truly do want
to recommend the most economical so-
lution. To do that, they need to under-
stand the process and control require-
MARCH 2017 49
 ments and the expected life cycle of the
unit where the offering will be deployed.
They can recommend the best mea-
surement technology or valve selection
and help size the instrument as well.
They can help evaluate the value and
return of additional options or choices.
So leverage their special knowledge and
expertise. Often on a larger project, an
engineering firm with its own subject-
matter experts may help with selection
and procurement, or selection may be
defined by corporate guidelines or pro-
cess licensing requirements.
In the final analysis, the control engi-
neer is trying to identify and understand
the sources of variability that can impact
product quality, production throughput,
yields, utility consumption and other
economic impacts, and tries to design
controls to attenuate the variability or
move it to a part of the process where
it has less economic impact. By simply
reducing variability, it is possible to op-
erate nearer to constraints and hence
maximize the processing capability of
the existing plant.
Chemical engineers who are inter-
ested in process control can find many
resources to gain a deeper understand-
ing in the further reading section. Q
Edited by Scott Jenkins
Further reading
1. Shinskey, F.G., “Process Control Systems: Application, Design
and Tuning,” 4th ed., McGraw Hill, 1996.
2. McMillan, G., “Handbook of Control Engineering,” McGraw Hill,
1999.
3. Blevins, T. and Nixon, M., “Control Loop Foundation – Batch
and Continuous Control,” International Society for Automation
(ISA), 2011.
Author
Lou Heavner is a control engineer at
Emerson (1100 West Henna Boulevard,
Round Rock, TX 78681; Phone: +1
(512) 834-7262; Email: Lou.Heavner@
Emerson.com). Heavner has been with
Emerson for over 30 years, with respon-
sibility for project engineering, sales and
consulting with customers across all of
the process industries and all over the
world. His current role is to scope and
lead advanced control and optimization projects, primarily in
the oil-and-gas industry. Prior to joining Emerson, Heavner
worked for Olin Corp. as a process automation and control
engineer. Heavner earned a B.S.Ch.E. degree from the Mas-
sachusetts Institute of Technology (MIT) in 1978. When he
isn’t helping customers control their processes, Heavner en-
joys controlling his own home brewery operation. He may be
contacted on LinkedIn at www.linkedin.com/in/louheavner.

WHERE THE INDUSTRY TURNS

FOR PRECISE CONTROL
Achieve precise control over every aspect of your chemical operation.

MODEL 521 MODEL 1049

The Model 521 is a sliding The Model 1049 Secure-Gard
stem, globe style, bellows is a pilot-operated vent valve
intended for installation
MODEL 1078 sealed, pneumatically
on atmospheric and low-
The Model 1078 actuated control valve
designed for maximum pressure storage tanks,
Vacu-Gard is a pilot- vapor recovery systems,
operated valve, corrosion resistance in pure
chemical service. and process systems.
specifically designed to
reduce blanketing gas
losses on low-pressure
storage tanks.
MODEL 3100
The Model 3100 is a
MODEL 987 pressure/vacuum vent
The Model 987 is a pneumatic control designed to vent the tank
valve designed to control moderate vapor away to atmosphere
to severe corrosive applications but and to relieve vacuum
may be applied in general service pressure within the tank.
applications also. The 3100 is a weight
loaded style.

www.cashco com
Cashco, Inc. Innovative Solutions

P.O. Box 6, Ellsworth, KS 67439-0006

Ph. (785) 472-4461, Fax: (785) 472-3539 www.cashco com
Innovative Solutions

Circle 11 on p. 90 or go to adlinks.chemengonline.com/66426-11
50 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017
 Environment
and Safety
V. CHADHA, S. BHANDARI and P. VIRDI,
Bechtel, Haryana, India

Control design for steam vent noise in refineries

Noise pollution in oil refineries has a direct effect on worker • Depressurization of steam lines for system maintenance
health. Noise is the unwanted or undesirable sound produced • Steam ejector discharge and exhaust.
by process control equipment, which includes valves. Sound is
produced by fluctuating pressure waves that arise from the flow NOISE ABATEMENT METHODS
of fluid through the valve. The specific sources are either me- Source treatments. Not only do source treatments re-
chanical or of fluid structure in nature. These pressure waves duce the strength of the noise source, but they also shift
are characterized by both amplitude (volume) and frequency. the frequencies to a higher, non-problematic range. Source
Amplitude is measured in decibels (dB). Industrial noise levels treatments prevent noise at its source, which can involve the
are often expressed in a form that takes into account human valve. Common source treatments include noise attenuation
sensitivity to frequency. This is known as “A-weighting” and is valve trims, and inline and external diffusers that minimize
measured in units called A-weighted decibels (dBA). turbulence. TABLE 1 provides a list of valve and diffuser de-
Many refineries have a specification for noise levels within signs that can be used, depending on their noise-reduction
the plant environment and, in some cases, they are subject capabilities.
to fence line noise requirements to protect the refinery’s sur- Path treatments. In addition to eliminating noise at its
rounding environment. As populated areas are moving closer source, path treatments can impede the propagation and trans-
to processing refineries, noise attenuation is crucial to adhere to mission of sound waves. Path treatments help eliminate noise
compliance requirements and avoid potential regulatory action. heard outside of the pipe wall by increasing the resistance of
Per ISA 75.171 or IEC 60534-8-3/60534-8-42,3, the maxi- the noise path. Typical path treatments include increasing the
mum permitted noise level, measured at 1 m from the valve, pipe thickness, adding acoustical or thermal insulation, or
must be less than 85 dBA. For all gas, vapor, steam and flashing adding inline pipe silencers.
liquid valves that have a typical pressure drop of 0.4 × upstream
pressure, a noise calculation must be supplied. If the predicted TABLE 1. Typical source treatment noise attenuations
noise level is found to exceed 85 dBA, then the valve will be Treatment Typical noise attenuation Typical application
treated for noise.
Single-stage cage trim Up to 18 dBA Steam letdown
The venting of high-pressure steam to the atmosphere dur-
ing plant operations often leads to unacceptably high noise lev- Multiple-stage cage trim Up to 30 dBA Gland steam
els. This can be a result of: External diffusers Up to 40 dBA Sky vent
• Steam safety relief valve discharge Inline diffusers Up to 50 dBA Steam letdown
• Dumping of steam during steam generator startup

Silencer Vent to atmosphere

Vent to atmosphere

Valve with diffuser assembly

Valve

Steam generator Steam generator

FIG. 1. Valve with a silencer assembly (path treatment). FIG. 2. Valve with a diffuser assembly.

Hydrocarbon Processing | DECEMBER 2016 73

 Environment and Safety
Silencers offer noise attenuation up to 45 dBA, depending
on treatment and application. The weakness of path treatments
is that both the fluid stream and the piping itself are very effec-
tive at transmitting noise; therefore, complete noise attenuation
Atmosphere
Tube
Baffle
Shell
Intermediate pressure
Atmospheric pressure
Inlet pressure
FIG. 3. Valve with a diffuser assembly.
NEW VERSION
InstruCalc
CONTROL VALVES • FLOW ELEMENTS • RELIEF DEVICES • PROCESS DATA
InstruCalc 9.0 calculates the size of control valves,
flow elements and relief devices and calculates fluid
properties, pipe pressure loss and liquid waterhammer
flow. Easy to use and accurate, it is the only sizing
program you need, enabling you to: Size more than 50
different instruments; Calculate process data at flow
conditions for 54 fluids in either mixtures or single
components and 66 gases, and; Calculate the orifice
size, flowrate or differential range, which enables the
user to select the flow rate with optimum accuracy.
Updates include Engineering Standard
Upgrades and Operational Improvements
in InstruCalc Version 9.0
Order Direct from the Publisher.
GulfPub.com/InstruCalc or call +1 (713) 520-4426.
74 DECEMBER 2016 | HydrocarbonProcessing.com
is difficult. This means that attenuation is only effective for the
area where the treatment is applied.
CASE STUDY
Initial design. A common design on various projects con-
sists of a steam generator with a valve and silencer assembly
for noise reduction (FIG. 1). When operated, the large pressure
differential across the sky vent valve can cause high noise lev-
els. To reduce the noise potential, the pressure drop is split be-
tween the valve and a downstream silencer. The pressure split
also serves to optimize the size of each component. Other key
challenges include:
• Valve leakage due to an inadequate seat load,
which could lead to reduced plant load and
efficiency, as well as damage to the valve
• Fast opening in case of a plant upset condition
• Properly addressing noise for two separate noise
sources (valve and silencer).
This design requires additional support requirements, an in-
crease of weight/size, and the need for a larger platform. This
type of design is more capital intensive.
Proposed design. All of these limitations were overcome by
using a combined assembly consisting of a valve with a diffuser
(FIG. 2). A diffuser is a pressure-reducing device that is installed
downstream from the valve. When installed, total pressure drop
of the system is divided across the valve and the diffuser. This
enables the valve to operate at a lower pressure drop ratio, there-
by lowering the noise level generated from the process flowing
through the valve.
At this location, the vent diffuser was used, which helped
reduce the noise generated by steam being vented to the at-
mosphere. The valve can be used during startup and abnormal
conditions. A diffuser valve combination can result in a noise re-
duction of up to 40 dBA, and retains the pressure/flow control
associated with a standard valve.
In the initial design valve and silencer proposed by the licen-
sor to attenuate noise, the valve and silencer are procured and
designed by different disciplines, which can lead to overdesign
and result in a higher cost. In the design proposed here, noise
measurement is considered in its entirety for the venting of
steam to the atmosphere. Different treatments are worked out
to reduce noise, which helps to achieve an optimized and cost-
effective design.
RECOMMENDATION
The assembly shown in FIG. 3 is a better and more cost-effec-
tive assembly than that shown in FIG. 1 to fulfill the requirements
of ISA 75.17 or IEC 60534-8-3/60534-8-4 maximum permit-
ted noise level. The valve diffuser assembly in FIG. 3 requires no
maintenance after installation. These features combine to offer
increased noise control at a minimum investment.
Although the case study discussed here comes from a refin-
ery application, this design may be implemented in any steam
service application.
LITERATURE CITED
1
ISA 75.17: Control valve aerodynamic noise prediction.
2
IEC 60534-8-3/60534-8-4: Noise considerations–Control valve aerodynamic
noise reduction method.
3
IEC 60534-8-4: Noise considerations–Prediction of noise generated by hydrody-
namic flow.
 LNG Technology
I. NEMOV, Sakhalin LNG, Sakhalin Energy
Investment Co. Ltd., Yuzhno-Sakhalinsk, Russia

Optimal control of mixed refrigerant

composition at Sakhalin LNG plant
A technique for the optimization of • The Carnot factor was used as on calculated component
mixed refrigerant (MR) composition was an invariant parameter to assess inventories rather than on
developed and successfully applied at the refrigeration cycle performance. component concentrations.
Sakhalin LNG plant on Sakhalin Island, The invariance is a crucial feature This transformation effectively
Russia. Indirect control of the MR com- of the factor that enables the linearizes and decouples the
position was implemented via calculated comparison of cycle performances system of controlled variables,
inventories, replacing direct control of in different plant operation cases. contributing to higher control
component concentrations and maximiz- • Clear correlations were observed performance and, thereby, more
ing the LNG plant’s performance. between the Carnot factor and the efficient plant operation.
MCHE temperature approaches,
Introduction. The MR cycle is used for as well as between the temperature Process description. The Shell DMR
liquefaction and subcooling of natural approaches and the concentrations process applied at the Sakhalin LNG
gas in most large LNG trains based on of MR components. From these plant uses a PMR cycle and an MR cycle
technologies such as AP-C3MR, AP-X correlations, functions of optimal to liquefy natural gas.1 The MR cooling
and Shell DMR. The varying composi- MR composition setpoints were cycle—in which the refrigerant is a mix-
tion of MR cycle technology is its ma- derived and implemented in the ture of nitrogen, methane, ethane, pro-
jor advantage over the Cascade process. plant’s control system. pane and traces of butane—provides the
This advantage comes from the fact that • To maximize control effectiveness, cooling duty for the liquefaction of gas
MR composition can be adjusted to fit a new MR composition control within a temperature range of –40°C to
the natural gas condensation curve and strategy was applied. It is based –150°C (FIG. 1).
maintain an optimum temperature ap-
proach between cold and warm streams
inside the main cryogenic heat exchanger
(MCHE) for a wide range of NG inlet
temperatures, depending on propane or
precooling MR (PMR) cycle duty. Addi-
tional degrees of freedom related to MR
composition help increase the efficiency
of the refrigerant cycle. The correct setup
of MR composition setpoints, as well as
their robust control, are key problems
that must be solved to maximize the ef-
ficiency of any MR cycle.
The MR composition optimization and
process control solutions described in this
article were developed and implemented
in-house at the Sakhalin LNG plant oper-
ated by Sakhalin Energy Investment Co.
Ltd. The research results were success-
fully applied on the MR circuit of the Shell
DMR process, but they are generally valid
for any MR cycle, irrespective of specific
LNG technology details. The results are
FIG. 1. MR cycle of Shell double mixed refrigerant process.
summarized in the following points:
Hydrocarbon Processing | JANUARY 2019 61
 LNG Technology
MR vapor exiting the shell side of the
MCHE’s E-07A/B section is compressed
in MR compressors K-01 and K-02. Inter-
cooling and initial discharge cooling are
achieved by air coolers E-01, E-02, E-03
and E-04, while further cooling and par-
tial condensation are achieved by cooling
against PMR at two different pressure
levels—the E-05 high-pressure PMR pre-
cooler and the E-06 low-pressure PMR
precooler. Cooled and partially con-
densed MR is separated into light mixed
refrigerant (LMR) vapor and heavy mixed
refrigerant (HMR) liquid in MR sepa-
rator vessel V-02. The HMR and LMR
high-pressure streams are subcooled sepa-
rately inside the MCHE and returned to
the shell side after expansion, providing
cooling duty for gas liquefaction.
The separation of MR in V-02 creates
two cycles with different compositions
and circulation rates. The composition
of refrigerant in different sections of the
MR cycle is far from uniform, which is a
challenge for tight control.
Changing process conditions for
two-phase MR mixture flashing in V-02
are the main source of variability in MR
composition measurements. The flash-
FIG. 2. Real (A) and ideal (B) refrigeration cycle diagrams showing compression with
inter-cooling and after-cooling in air coolers (points 1–2); MR cooling in the PMR cycle
(points 2–3); MR self-cooling in the MCHE (points 3–4); Joule–Thomson expansion (points 4–5);
and MR boiling in the MCHE (points 5–1).
FIG. 3. Correlations between the Carnot factor of the MR cycle and the MCHE temperature approaches (Tcp slice from –47°C to –48°C).
62 JANUARY 2019 | HydrocarbonProcessing.com
ing is defined by three parameters:
• The MR temperature in the E-06
outlet, which depends on PMR
cooling duty
• The LMR to HMR mass ratio
(LMR/HMR), which is maintained
by the control system
• The MR composition, which
depends on the V-02 separation
temperature and LMR/HMR,
as well as on component
inventories in the MR system.
The pressure in V-02 results from
vapor-liquid equilibrium. A two-phase
mixture with a given composition at a
given temperature splits the vapor and
liquid phases with a given ratio between
them, but only at a certain pressure. In
this system of mutually dependent pa-
rameters, the MR component concentra-
tions are subject to the influence of many
factors, and the component inventories
are the only invariants. In general, the
inventory of an MR component in the
system can be changed in a controlled
way by adding a certain component and
removing either the HMR mixture or the
LMR mixture. Changes in pressure and
temperature in different parts of the MR
circuit affect the MR holdup in those
parts, while the total inventory is fixed
and the HMR liquid level in V-02 serves
as a buffer for those fluctuations.
Optimization of MR composition
setpoints. Most MR composition opti-
mization techniques deal with an offline
process model, rather than with a running
plant.2,3 Usually, such techniques repre-
sent a combination of a relatively simple
model of a liquefaction unit and a fully or
partially automated optimization routine.
However, at Sakhalin LNG, this meth-
od was found not to work well for the opti-
mization of a running plant. Any mismatch
between process equipment models and
actual plant operation resulted in a sub-
optimal solution where the actual plant
performance did not reach the predicted
level, or even decreased. The adoption of
a rigorous, model-based MR optimization
of a running LNG plant has potential, but
this depends on the further development
of modeling techniques.
Statistical treatment of past operational
data is the most prevailing method adopt-
ed by the LNG industry for the optimiza-
tion of MR composition at running plants.
It provides credible results if the data
population is large enough to represent
the variations of MR composition and
cycle efficiency. The method is based on
the extraction of top efficiency data from
the total data population, as well as the de-
velopment of correlations between PMR
or propane cycle cooling temperature and
optimum MR component concentrations.
The main challenge of this method is de-
riving an invariant performance indicator
that is sufficient to describe cycle efficien-
cy. The estimation and application of the
Carnot factor for MR composition opti-
mization at Sakhalin LNG is described in
the following paragraphs.
If PMR cycle capacity is given and
fixed, the total flowrate of LNG measured
 LNG Technology

just downstream of the MCHE’s E-07A/B • The hot reservoir temperature verted into an analytical form as a func-
unit depends on several factors: and the gas inlet temperature tion of temperature (Eq. 7):
• Gas temperature on the E-07A/B are the same. dqNG = λ(T) × dT (7)
inlet (i.e., cutpoint temperature, The minimum specific work of the where:
or Tcp) as defined by PMR cycle ideal cycle can be derived from the basic λ(T) = Natural gas specific heat func-
operating conditions principles of the Carnot cycle3 (Eqs. 3, tion of absolute temperature.
• LNG temperature on the E-07 4 and 5): Finally, after integration of Eq. 6, mini-
A/B outlet (i.e., LNG rundown dwid = dqPMR − dqNG (3) mum specific work Wid can be expressed
temperature, or Trd) as defined as a function of Tcp and Trd and then used
by the control system dqPMR = Tcp × dS (4) for Carnot factor estimation (Eq. 2), as
• The available power of MR shown in Eq. 8:
dqNG = T × dS (5)
compressor drivers (KT-01 λ(T)
and KM-01) where: Wid = ∫dwid = Tcp × ∫ TTrdcp (8)
T
• MR cycle coefficient of NG = Natural gas
performance, which is the ratio of dqNG, dqPMR = Amount of heat removed × dT − ∫ TTrdcp λ(T) × dT
cooling duty provided to the gas from the gas stream and transferred to a
stream over the compressor drivers’ high-temperature reservoir (PMR cycle)
total power. It can vary depending dS = Entropy change corresponding to
on MR cycle process conditions, as dqNG heat removed from the gas stream at
well as on equipment availability. absolute temperature T.
The actual performance coefficient of Combining Eqs. 3, 4 and 5, Eq. 6 can
the MR cycle can be calculated as shown be obtained in differential form for ideal
in Eq. 1: specific work:
FLNG × ∆H LNG dq
COPact = (1) dwid = (Tcp − T) NG =
PMR T
(6)
where: dqNG
Tcp × − dqNG
FLNG = LNG flowrate T
∆HLNG = NG enthalpy change over the Condensation heat removed from the FIG. 4. MCHE temperature approaches
temperature range from Tcp to Trd gas can be obtained in numeric form from corresponding to top 15% efficient operation
PMR = Power of MR drivers delivered a thermodynamic package and then con- vs. historical operation data variance.
to MR compressors.
The MR cycle coefficient of perfor-
mance is a value invariant to the changes
of available power, since the LNG pro-
duction rate is a function of the power
(1). The ratio of actual COP to ideal
COP (Carnot factor) was used to obtain
a value that is also invariant to Tcp and Trd,
as shown in Eq. 2:
COPact FLNG × ∆H LNG
ψ= = ×
COPid PMR (2)
Wid FLNG × Wid
=
∆H LNG PMR
where:
Wid = Minimum specific work/1 kg of
gas required to cool it from Tcp to Trd in an
ideal refrigeration cycle.
Two assumptions were made when de-
fining an ideal refrigeration cycle for gas
liquefaction (FIG. 2):
• The ideal cycle can be represented
as the number of Carnot cycles,
each operating between a specific
cold reservoir temperature
(gas condensation curve) and
a hot reservoir temperature, FIG. 5. MR composition corresponding to the optimum MCHE temperature approaches (error
which is common for all the cycles bars correspond to standard deviations of the correlations) vs. historical operation data variance.

Hydrocarbon Processing | JANUARY 2019 63

 LNG Technology

A major advantage of the Carnot fac-
tor is its invariance, which allows for the
comparison of MR cycle performance
based on a single scale, irrespective of the
PMR cycle capacity, the available power
of MR compressor drivers and varia-
tions in LNG rundown temperature.
Therefore, the yearly probability dis-
tribution of the Carnot factor can help
estimate the available room for cycle
efficiency increases, as well as confirm
positive effects after the implementation
of optimization-related changes. Several
operations were performed on historical
operation data, including Carnot factor
values array:
• Instrumentation bad data and
periods of turndown or transient
operation were filtered out, along
with equipment unavailability and
malfunctions
• Carnot factor inaccuracies due to
instrumentation noise, as well as
natural variance from suboptimal
plant operation, were estimated
• The maximum possible increase in
MR cycle efficiency was estimated
• Data was checked for correlation
between several process parameters
and the Carnot factor
• Functions of optimal MR
composition setpoints were
derived based on top-performance
operational data
• Derived optimum functions were
validated by applying them to all
data populations and estimating an
expected economic impact.
The list of process parameters checked
for correlation with the Carnot factor in-
cluded the MCHE’s superheat margins
and temperature approaches, the MR
compressor pressure ratio, the LMR/
HMR mass flow ratio, the LMR and
HMR compositions, the MR component
concentrations and inventories.
Clear correlations were found only
between the MCHE temperature ap-
proaches and the Carnot factor, and
between the MR component concentra-
tions and the MCHE temperature ap-
proaches. This observation is important,
since it confirms the theoretical conclu-
sion that MR cycle efficiency strongly
depends on MCHE temperature ap-
proaches (FIG. 3). At the same time, real
plant data indicate that the minimum
possible temperature approach does not
correspond to the highest efficiency,
likely because of limited heat transfer
area and developing temperature pinch.
Moreover, it was found that MR compo-
nent inventories do not correlate with
the cycle efficiency. This is the reason
why the inventories cannot be used in
an MR composition control strategy as
self-sufficient optimization targets. In
other words, a wide range of component
concentrations (including optimum and
sub-optimum sets) can correspond to a
single set of component inventories.
Operational data was sliced by Tcp val-
ues with a 1°C interval, and the slices were
treated to extract cases with a high Car-
not factor (top 15%) with corresponding
MCHE temperature approaches. The av-
erage values of extracted temperature ap-
proaches were used to derive optimal ΔT
functions (FIG. 4). Finally, the functions
of optimum temperature approaches
were used to extract the corresponding
MR component concentrations data and
build correlations between Tcp and opti-
mum MR composition (FIG. 5). Those
correlations could then be used for con-
tinuous scheduling of MR composition
setpoints in a control system.

Indirect MR composition control

via component inventories. Statis-
FIG. 6. Schematics of MR direct concentrations control. tical treatment of operational data and
MR cycle efficiency provides valuable
information about minimum MR com-
TABLE 1. Multiple input/multiple output (MIMO) model structure for direct position control compliance required to
concentration control achieve the expected increase in perfor-
Controlled variables mance. Standard deviations of MR com-
C1 C2 C3 N2 L V-02 P V-02
ponent concentrations corresponding to
the top-performance operation from the
C1 makeup ++ – – – + +
Manipulated variables

derived functions of MR optimum com-

C2 makeup – ++ – – ++ – position were used as control compli-
C3 makeup – – ++ – ++ – ance benchmarks (see error bars on the
N2 makeup – – – ++ + ++
trends in FIG. 5). This makes sense, since
composition control tightness must be at
HMR drain – –– –– – – 0
least as good as the correlations between
LMR vent –– – – –– 0 – MR composition and cycle efficiency.

64 JANUARY 2019 | HydrocarbonProcessing.com
 LNG Technology

Maintaining MR component concentra-
tions within a ±0.3% target window is a
challenge, since measured MR composi-
tion continuously experiences disturbance
effects from several factors:
• Adjustment of LMR/HMR by
control system
• Variation of PMR capacity resulting
in Tcp changes
• Adjustment of MR compressor
suction pressure by control system
• Variation of MR circulation rate
• Possible MR composition drift
due to passing valves and leaking
MCHE tubes.
The most straightforward method of
MR composition control relies on the di-
rect use of component concentrations in
the control scheme (FIG. 6). The V-02 level
MR component concentrations into actual form refrigerant composition across the
and target inventories involves the input circuit (varying gas compressibility and
of V-02 level measurement and the level liquid density), as well as the presence of
setpoint. It results in the indirect control two-phase mixtures in some sections of
of not only MR composition, but also the the MR circuit.
V-02 level. V-02 temperature changes are
realized in smaller variances in V-02 pres- m Pi × Vi
n j = R × ∑i=1 Cij × (9)
sure, due to the higher performance of the Zi × Ti
inventory control scheme; therefore, the where:
use of a V-02 pressure protective control nj = Amount of substance of j compo-
loop can be avoided. nent in mixture
While refrigerant composition control
is traditionally implemented with the use
of multivariable predictive control tech- TABLE 2. Multiple input/single output
niques, the decoupling of the controlled (MISO) model structure for inventory
control
variables provides more freedom in the se-
lection of control technology. An example Controlled variables
of a simplified control scheme for the two- C1_i C2_i C3_i N2_i
component system is shown in FIG. 7.
C1 makeup + 0 0 0

Manipulated variables
and pressure protective controls must be While pressure and temperature condi-
involved in that scheme, since the level tions and volumes are known for most of C2 makeup 0 + 0 0
does not depend on concentrations but the MR circuit sections, the continuous C3 makeup 0 0 + 0
rather on a measure of component inven- online estimation of component invento- N2 makeup 0 0 0 +
tories and must be controlled separately. ries accumulated in the refrigerant loop is
HMR drain – –– –– –
At the same time, variations in the pressure still not a straightforward operation (Eq.
allowed by that control scheme must be 9). Complexity comes from the non-uni- LMR vent – – – ––
moderated by protective control.
A general analysis indicates that no con-
trollability issues are anticipated, since the
numbers of controlled and manipulated
variables are equal and each component is
independently controlled (TABLE 1). How-
ever, several crucial features still limit the
control performance of this scheme:
• The ramp character of process
models between all manipulated
and controlled variables
• Strong mutual dependency of
the controlled variables
• Non-linear gains, depending on
LMR and HMR composition
• N-1 degrees of freedom in the system
of N component concentrations,
which make independent control FIG. 7. Simplified example of base-layer inventory control scheme for a two-component system.
impossible.
The combination of listed features
makes tight control impossible with the de-
scribed scheme.
From a process control point of view, the
inventory of each component is more suit-
able for control than its concentration in
the MR mixture. The inventories are invari-
ant to process changes (Tcp, LMR/HMR,
etc.) and completely decoupled from one
another, which simplifies the model struc-
ture from multiple input/multiple output
(MIMO) to multiple input/single output
(MISO) (TABLE 2).
FIG. 8. Schematics of MR inventory control.
The transformation of actual and target

Hydrocarbon Processing | JANUARY 2019 65

 LNG Technology
FIG. 9. Calculation schematics for gas-liquid equilibrium in V-02 for actual and target MR
concentrations.
FIG. 10. MR composition offset and variation before and after the implementation of an
inventory control scheme based on plant operational data.
R = Universal gas constant
Cij = j component concentration in i
section of MR system
Pi, Ti = Pressure and temperature in
the i section
Vi = Volume holdup of the i section
Zi = Compressibility of real gas in the
i section.
A common solution for the outlined
problems is to use the equation of state
(EOS) solver for the continuous estima-
tion of gas, liquid or two-phase mixture
properties in all MR cycle sections. This
approach was successfully applied at the
Sakhalin LNG plant to estimate MR com-
ponent inventories corresponding to ac-
tual and target concentrations.
The Peng-Robinson EOS solving algo-
rithm4 was implemented at the distribut-
ed control system (DCS) as a structured
text program developed in-house. It is ro-
bust against abnormal process conditions,
like quick changes of parameters or trips,
as well as instrument signal abnormalities.
Several estimators were implemented
for different sections of the MR circuit,
linked with corresponding inventory
66 JANUARY 2019 | HydrocarbonProcessing.com
calculations based on volume holdups
of equipment and pipelines. The calcu-
lations provided estimates of actual and
target inventories across the MR circuit,
which are used for control (FIG. 8).
V-02 temperature is assumed to be
the same for the actual and target con-
ditions, since the temperature is an ex-
ternal parameter depending only on the
PMR cycle cooling capacity. Pressure
and flash rate, in combination with tem-
perature and mixture composition, com-
pletely define two-phase equilibrium. In
the inventory control of MR separation,
flash rate is a fixed variable since it is
directly related to the controlled LMR/
HMR; therefore, the calculation of EOS
for target conditions is supplied with an
LMR/HMR target value that matches
the operating conditions, and the pres-
sure is adjusted to match the ratio. Note:
The calculation of gas-liquid equilib-
rium in the V-02 vessel must be time-
synchronized for actual and target MR
compositions, as shown in FIG. 9. This
ensures that calculation outputs will be
delivered to inventory estimates at the
same time and will correspond to the
same process parameters. This point is
crucial for minimizing the effect of mea-
surement noise on parallel actual and
target inventory estimates.
The implementation of MR inventory
control has delivered excellent results in
terms of MR composition compliance
with optimum targets. An analysis of one
month of post-commissioning opera-
tional data showed that MR composition
had been maintained inside the specified
target windows for approximately 98.5%
of controller uptime, while the windows
were specified at ±0.3 mol%, on average
(the correlation error bars in FIG. 5 refer to
the specified target windows).
The average offset between MR com-
ponent concentrations and dynamically
set targets was reduced to zero, and the
variance of the error was decreased by sev-
eral times (FIG. 10). The inventory control
scheme was also found to be sufficiently
robust to work autonomously, without
any adjustments by the control room op-
erator, as opposed to the previously used
direct-concentration control scheme that
required 8–10 manual operator actions
per day. These results show that inventory
control is the correct strategy for improv-
ing MR composition control, resulting in
increased LNG plant efficiency.
The methods of MR composition opti-
mization and control described here have
been patented in Russia (RU 2640976
C1). Similar patent applications have been
submitted in six other countries (Austra-
lia, Indonesia, Malaysia, Nigeria, Qatar
and the US) and are under review. GP
LITERATURE CITED
1
Dam, W. and S.-M. Ho, “Unusual design consider-
ations drive selection of Sakhalin LNG plant facili-
ties,” Oil & Gas Journal, Vol. 40, 2001.
2
Khan, M. S., et al., “Knowledge-based decision-mak-
ing method for the selection of mixed-refrigerant
systems for energy-efficient LNG processes,” Applied
Energy, Vol. 111, 2013.
Complete literature cited available online at
GasProcessingNews.com.
IVAN NEMOV is a Senior Process
Control Engineer who worked at
Sakhalin Energy Investment
Company Ltd. from 2012–2017,
mainly supporting Sakhalin LNG
plant operation. In 2017, he joined
Shell Australia Pty. Ltd. for
commissioning and startup of the Prelude FLNG vessel.
He earned his ME degree from Gubkin Russian State
University of Oil and Gas in Moscow, Russia in 2012 and
teaches as an Assistant Professor at Astrakhan State
Technical University in Astrakhan, Russia.
 Cover Story
Part 1

Implementing an
‘Integrity Operating
Window’ Program
An effective Integrity Operating Window (IOW) program — which establishes safe operating
limits and acceptable limits of process variation before an asset begins to degrade — can help
operators stay ahead of potential repairs and reduce risk
Russ Davis System Description Color API 570 Corrosion Damage Limits
MISTRAS Group Pipe Class rate, mil/yr mechanism
100 Ammonia vapor 1  O.5 NH3 SCC* Carbon steel in aqueous
ammonia non-PWHT*
IN BRIEF 101 Ammonia liquid 2  1 Ammonium Concentration of ammonium
DEFINING INTEGRITY chloride cor- salts and temperature
rosion
OPERATING WINDOWS
102 Ammonia, CO2, H2O 2 3  NH3 SCC Carbon steel in aqueous
IOW PROGRAM- vapor, ammonia recycle ammonia non-PWHT
DEVELOPMENT High-pressure streams
OVERVIEW 300 Tail gas 2 3  CO2 corrosion <300°F
DAMAGE OR CORROSION Liquid phase where CO2
ANALYSIS condenses from vapor
phase
RISK RANKING
500 Natural gas 2  O.5 CO2 corrosion <300°F
EVALUATING ALARMS Liquid phase where CO2
condenses from vapor
DEFINING THE IOW
phase
CRITICALITY
501 Fuel gas 2 O.5  CO2 corrosion <300°F
DOCUMENTATION AND
Liquid phase where CO2
TRAINING condenses from vapor
phase
IOW INTEGRATION
502 Hydrogen line 2 O.5  CO2 corrosion <300°F
Liquid phase where CO2
condenses from vapor
phase

A
comprehensive asset-integrity
management (AIM) program is es-
sential to ensuring process integ-
rity and reducing risk of operations
to as low a level as practicable. Historically,
these programs are based on a given set
of operational parameters that were es-
tablished during the implementation of
the program. To develop these programs,
process conditions — such as the various
chemical components present in the pro-
cess streams, temperature, pressure, and
flowrates — are analyzed by mechanical in-
42 CHEMICAL ENGINEERING
FIGURE 1. Damage mechanisms are established to identify
the key controllable monitoring parameters, or variables that
can be effectively adjusted to bring the process back within
safe operating limits (*SCC = stress corrosion cracking;
PWHT = post-weld heat treatment)
tegrity (MI) subject matter experts (SMEs).
These experts identify the potential damage
mechanisms associated with anticipated
operating conditions, and help to establish
inspection methods and testing frequen-
cies. In the subsequent management of the
AIM program, reliability and MI groups rely
on these parameters and testing method-
WWW.CHEMENGONLINE.COM MAY 2017
ologies that were established during the requires a multi-disciplinary team of engi-
program’s development. neering SMEs, with advanced knowledge of
But over time in a process facility, op- mechanical, corrosion, reliability and mate-
erational parameters change, and the rate rials engineering disciplines; a multitude of
of equipment degradation may change as inspection and maintenance services; and
well, causing the asset-integrity program to plant data-management software to store
become ineffective. This can shorten equip- and trend data over the program’s lifetime.
ment lifespans, raise costs, and greatly en- With such a wide range of SMEs and com-
hance potential risk to personnel, assets and petencies required, operating companies
the environment. may partner with an experienced third-party
To maintain the integrity of process equip- service provider with expertise in the full
ment in petroleum refining and petrochemi- scope of asset-protection solutions, includ-
cal facilities, a comprehensive process safety ing inspection, engineering, maintenance,
management (PSM) system is required. Cur- condition-monitoring, and mechanical in-
rently, most of these systems are oriented tegrity data management software. Through
toward a rigorous mechanical integrity pro- such a partnership, the stakeholders can
gram that predicts or prevents failures of develop, implement and manage an effec-
pressure equipment. Beyond this, however, tive program.
the optimal PSM support systems include
implementation of so-called integrity oper- Defining integrity operating windows
ating windows (IOWs; defined below) and As defined by the American Petroleum Insti-
management-of-change (MOC) programs, tute’s (API) RP 584, integrity operating win-
which are designed to monitor and control dows are established limits for process vari-
process variables that can impact the likeli- ables (parameters) that can affect the integrity
hood of failure. of the equipment if the process operation
The implementation and management deviates from the established limits for a pre-
of an effective IOW program is complex. It determined length of time [1]. In short, IOWs

T +49 2961 7405-0 | info@rembe.de

Your Specialist for

PRESSURE RELIEF SYSTEMS
and
© REMBE | All rights reserved

EXPLOSION SAFETY

Consulting. Engineering. Products. Service.

Gallbergweg 21 | 59929 Brilon, Germany

F +49 2961 50714 | www.rembe.de

Inc.
3809 Beam Road Suite K Charlotte, NC 28217, USA
T +1 704 716 7022 | F +1 704 716 7025
info@rembe.us | www.rembe.us

Circle 35 on p. 90 or go to adlinks.chemengonline.com/66428-35 Circle 05 on p. 90 or go to adlinks.chemengonline.com/66428-05

CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MAY 2017 43
 Cover Story
FIGURE 2. As shown in this
sample process flow dia-
gram, process flow streams
and control groups are often
color-coded in accordance
with the colors assigned to
them on the damage-mecha-
nism analysis table shown in
Figure 1
44
establish acceptable limits of process varia-
tions before an asset begins to incur damage.
The purpose of defining an IOW is to es-
tablish, implement and maintain a program to
identify any potential damage mechanisms
that may adversely affect the process, and
then use that information to create a system
where parameters can be modified as pro-
cesses evolve over time. IOWs identify safe
operating limits with the express purpose of
avoiding equipment degradation that could
lead to a loss of containment. The ultimate
goal of IOWs is to lower the risk of operating
plant process equipment.
Integrity operating windows are typically
defined by their criticality. API RP 584 divides
IOWs into three categories:
1. IOW critical limit
2. IOW standard limit
3. IOW informational limit
IOW critical limit. An IOW critical limit would
be one that, if exceeded, could lead to rapid
deterioration of process equipment. A criti-
cal limit requires immediate operator action
to return the process variable to a predeter-
mined parameter in order to prevent poten-
tial equipment damage in a short time frame.
An example of a critical limit could be a pH
excursion in a process stream, as an ex-
tremely low pH could quickly damage pres-
sure equipment and could reasonably lead
to loss of containment in a short time.
IOW standard limit. An IOW standard limit
CHEMICAL ENGINEERING
is an integrity parameter, which, if exceeded
over a specific time frame, could cause in-
creased corrosion rates or eventually lead
to cracking or other damage to materials
of construction. Standard limits are typically
very time-based, in that the time required for
equipment to be adversely impacted gener-
ally defines the response to a standard limit
IOW. The consequence of damage to the
equipment associated with the IOW will also
influence an SME’s response to the param-
eter being exceeded. An example of a stan-
dard limit IOW could be an elevated temper-
ature on a heater tube skin, which could lead
to tube failure over time.
IOW critical limit. IOW informational lim-
its are generally used by SMEs to predict
the long-term integrity of equipment, or to
analyze impact on the asset-integrity pro-
gram. Informational IOWs are parameters
that may or may not be affected by opera-
tions. These IOWs do not typically require
operational responses but may be utilized
to assess future repairs of turnaround and
shutdown frequencies. Informational IOWs
can be used to evaluate the process as-
sumptions used to establish risk-based in-
spection (RBI) programs. An example of an
informational limit would a temperature rise
due to process creep, which could indicate
to facility operators that the parameters that
were defined during the AIM program
development are misaligned with current
WWW.CHEMENGONLINE.COM MAY 2017
 operating processes.
These terms vary across compa-
nies and industries. Critical limit IOWs
may also be referred to as safe oper-
ating or safety critical limits. Standard
limit IOWs have been referred to as
key operating limits or reliability limits,
and informational limit IOWs may also
be referred to as corrosion control
limits, depending on the industry in
which they are used.
IOW program-development
overview
Step 1. The first step to developing
an IOW program is to identify all po-
tential damage or corrosion mecha-
nisms that may adversely impact the
process equipment. A diverse team
of engineers will evaluate the oper-
ating parameters of each process
condition to understand their po-
tential corrosivities, and will evaluate
their impacts on equipment materi-
als of construction. The SME team
will develop a damage or corrosion
map, which will be used again later
to evaluate what indicators are cur-
rently provided to detect exceed-
ance of an IOW limit. SMEs can then
define the IOW limits for each dam-
age mechanism.
Step 2. After all damage mecha-
nisms have been identified, the con-
sequence of failure and the likelihood
of failure need to be analyzed and
understood. Consequence of failure
data can be gathered from process
hazard analysis (PHA) data or by
consequence modeling. The equip-
ment failures are then risk-ranked.
Step 3. The third step in IOW pro-
gram development is typically an
evaluation of alarms, indications, and
procedures necessary to recognize
exceedance of the IOW limit.
Step 4. Step four is to define the crit-
icality of the operating limit and de-
fine the priority of the IOW limit. This
is where SMEs determine whether
the IOW is critical, standard or
informational.
Step 5. The next step is the docu-
mentation of each IOW, and the de-
velopment of proper responses to
IOW alarms and notifications. Pro-
cedures or work instructions should
define the roles and responses nec-
essary to return the process to a reli-
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
ably safe state, or to further assess plant operations, maintenance and
the response to limit exceedance, reliability programs, and plant data-
depending on criticality. management software. This is an es-
Step 6. After IOWs and procedures sential step, to ensure that IOW pro-
have been properly documented, all gram procedures are being uniformly
personnel involved in the process practiced throughout a facility, and
operation must be trained in IOW im- that any change to a process param-
plementation and timely responses eter is being catalogued alongside
to IOW indicators and alarms. the rest of the facility’s integrity data.
Step 7. Step seven is integrating Step 8. Finally, the last step to an
the IOW program into the rest of the IOW program is the revalidation of
YOUR COMPETITIVE EDGE IN
PROCESS HEAT TRANSFER TECHNOLOGY
Do you need accurate heat exchanger technology for real-world situations?
Software and services to help you design equipment that performs under the
broadest range of operating conditions? IF SO, YOU CAN RELY ON US.
HTRI is a leading provider of
products and services that help
engineers around the world rate, RESEARCH
simulate, design, and enhance the
performance of heat exchangers.
SOFTWARE
Our extensive expertise is the result
of more than 50 years of applied
research conducted in our
industrially relevant testing facilities. TECHNICAL
SUPPORT
We deliver research, proprietary
contract services, training, support,
and software, including our
acclaimed Xchanger Suite®, to PROPRIETARY
assure our customers the highest CONTRACTS
operating confidence in equipment
designed using our technology.
TRAINING
Visit www.htri.net
to learn more.
Circle 21 on p. 90 or go to adlinks.chemengonline.com/66428-21
MAY 2017 45
 Cover Story
Consequence category ure for the equipment within each system.
A B C D E
Consequence-of-failure data may come
< $10K $10K to $100K $100K to $1M $1M to $10M > $10M from process-hazard analysis (PHA) data
5 or consequence modeling, and analysis
0 0 0 2 0 may be performed as an element of the
1E to 1
4 risk assessment. Once a risk analysis is
0 0 0 2 0 completed, ranking can be performed and
1E-2 to 1E-1
3 the risk rank associated with each system
0 0 0 0 0 or subsystem can be documented.
1E-3 to 1E-2
1 The risk-ranking data will be used to cat-
0 0 0 1 0
0 to 1E-4 egorize the IOW limits, as shown in Figure 3.
Target risk 4 4 3 2 2 High-risk events will require a Critical IOW.
Events with medium risk may require a Stan-
dard IOW, and low-risk events may be cat-
Risk rank Low Medium Medium high High
egorized as requiring an Informational IOW
Bold line indicates target risk only.

FIGURE 3. Risk-ranking data
can be used to categorize the
IOW limits, as shown here.
High-risk events will require
a Critical IOW. Events with
medium risk may require a
Standard IOW, and low-risk
events may be categorized
as requiring an Informational
IOW only
46
the IOWs, which consists of reviewing each
IOW for effectiveness and avoidance of spu-
rious alarms and notifications.
Damage or corrosion analysis
A comprehensive analysis must be per-
formed by an engineering team with knowl-
edge of the relevant processes, operating
parameters, corrosion analysis and damage
mechanisms. This team assesses poten-
tial damages that can adversely affect the
equipment based on the material of con-
struction. They will also assess the operating
parameters of process conditions, such as
temperature, stream constituents, pressure,
vibration, abrasiveness and more. Once
the team has identified potential corrosive
streams and the equipment materials of con-
struction, the limits associated with the dam-
age mechanism are identified. The team may
have an output report such as that shown in
Figure 1.
Once the potential damage mechanisms
associated with the various process systems
and limits have been identified, these data
can be depicted on plant process flow dia-
grams (PFDs), such as the one shown in Fig-
ure 2. Process flow diagrams typically con-
tain all major pieces of process equipment,
identified by a unique number; all process
flow streams, identified by a number and
their chemical compositions; and control
loops, or groups of equipment and piping
with similar materials, operating conditions
and degradation mechanisms.
Risk ranking
Risk profiles should be developed for each
equipment item that has a potential influ-
ence on process safety. The risk-ranking
process consists of analysis of the conse-
quence of failure, and the probability of fail-
CHEMICAL ENGINEERING
Evaluating alarms
By understanding the limits associated with
potential damage mechanisms that may af-
fect the equipment, SMEs can perform an
evaluation of what alarms, alerts and notifi-
cations will be required for operations, reli-
ability and mechanical integrity groups to
recognize exceedance of an IOW limit.
Alarms are the typical indicators for criti-
cal limit IOWs, usually coming in the form of
horns and flashing lights in the control room
to denote that immediate action must be
taken. Alerts and notifications can span from
visual or audio signals to simple emails to
operations and technical personnel, inform-
ing them that an IOW has been exceeded
but without any urgent need for time-sensi-
tive action to be taken.
The engineering team typically assesses
process parameters such as the following:
• Temperature
• Pressure
• Flow
• Stream constituents
• Water content
• Chlorides
• Sulfur
• pH
Defining the IOW criticality
At this step, operators must determine IOW
criticality limits and priorities, depending on
potential damage severity and the expected
time constraints before serious damage oc-
curs.
• Critical IOW — An alarm requiring a timely
response by a facility operator or SME
to bring the process back within IOW
parameters
• Standard IOW — Typically includes an
alert to operations personnel and to the
reliability SME. Standard IOWs usually
WWW.CHEMENGONLINE.COM MAY 2017
 The ultimate for
small tanks!
The future is 80 GHz: a new generation
of radar level sensors

When it comes to contactless level measurement of liquids in

small containers, smaller is better. With the smallest antenna of
its kind, VEGAPULS 64 is simply the greatest! With its excellent
focusing and insensitivity to condensation or buildup, this new
radar sensor is truly exceptional. Simply world-class!

www.vega.com/radar

Wireless adjustment via Bluetooth with

smartphone, tablet or PC. Compatible retrofit to
all plics® sensors manufactured since 2002.

Circle 45 on p. 90 or go to adlinks.chemengonline.com/66428-45
 Cover Story
have a timeframe associated with them.
If an integrity limit is exceeded for a set
time, equipment will suffer damage
• Informational IOW — Information con-
veyed from field-gathered data to the
reliability SME. This information may
require changes to the RBI assumptions,
frequency of inspection, or nondestructive
evaluation (NDE) methodology
considerations
Documentation and training
Just like every other aspect of a process
safety program, documentation is critical.
The IOW program must include documenta-
tion of each IOW and the proper response
to IOW alarms, alerts and notifications. This
information should be incorporated into the
facility’s operating procedures as well as reli-
ability program documentation.
When process changes occur, the asset-integrity
group may not always be informed, or may be using
an outdated frame of reference from the program’s
original development.
Once all necessary IOW information has
been documented, the facility must properly
define the roles and responsibilities for op-
erations, engineering, reliability and mechani-
cal integrity personnel. All of this information
should be proceduralized, and all responsible
personnel should be trained in IOW imple-
mentation and the proper, timely response to
IOW alarms, alerts and notifications.
IOW integration
The IOW program must be fully integrated
into the plant’s operations and maintenance,
reliability and mechanical integrity programs,
and engineering processes. Changes in feed-
stock, temperatures and flow characteristics
that impact individual assets can have an im-
pact on the entire asset-integrity program. It
is critically important for a facility to fully un-
derstand the information returned from the
IOW program and to recognize the effects
that changes in one process parameter can
have on other equipment, as this information
should inform operations from purchasing all
the way through post-production.
This step can be time-consuming and
meticulous. In order to maximize its effec-
48 CHEMICAL ENGINEERING
tiveness, an IOW program must be as in-
tegrated into plant operations as possible.
Utilizing a comprehensive mechanical-
integrity data-management software pro-
gram, in which inspection, corrosion, and
integrity data, MOC activities, RBI man-
agement and more are stored, organized,
trended and analyzed — helps to ensure
that IOW program data are centrally located
and consistently updated.
Closing thoughts
Getting the longest, safest reliable life from
each equipment component in the plant is
critical to the profitability of any chemical
process operation. Process safety is an ev-
ergreen management program, and changes
that affect the program design must be mon-
itored and properly responded to in order to
maintain a risk rate that is as low as practi-
cable.
When process changes occur, the asset-
integrity group may not always be informed,
or may be using an outdated frame of ref-
erence from the program’s original develop-
ment. For example, a slight temperature ele-
vation can raise a corrosion rate significantly.
Changes in sulfur content in a crude stream
can lead to sulfidation damage in areas the
MI group had not anticipated.
The key for any facility is to control these
variables before they start harming equipment
and people. An effective IOW program — one
that is fully integrated into plant processes,
and implemented and managed by properly
trained personnel — helps facilities to be in-
formed when variables become unsafe, and
lets them know the actions that must be taken
to quickly reduce risk. An integrated IOW pro-
gram is an essential component of any me-
chanical integrity program. Q
Edited by Suzanne Shelley
References
1. American Petroleum Institute, Integrity Operation Windows, RP 584,
1st Ed. May 2014.
2. Center for Chemical Process Safety (CCPS), Guidelines for Asset
/
Integrity Management, John Wiley & Sons, Inc., Hoboken, New
Jersey, 2017.
Author
Russ Davis, CSP, is the National Asset Integrity Management Ser-
vices (AIMS) & Mechanical Integrity (MI) Center of Excellence Man-
ager for MISTRAS Group, Inc. (4000 Underwood Rd., La Porte, TX
77571; Phone: 281-478-1636; Email: russ.davis@mistrasgroup.
com). Davis has several decades of experience in program design
and implementation in the mechanical-integrity and asset-reliability
industries. He has extensive experience in process safety manage-
ment (PSM) and hazard assessment and consequence analysis.
Davis has also served as both an internal and external consul-
tant, directing the implementation of MI programs for companies
throughout the global CPI.
WWW.CHEMENGONLINE.COM MAY 2017
 Engineering Practice
Key Considerations in Specifying Control Valves
Follow this guidance to specify a control valve accurately during the design phase
Satyendra Kumar Singh
Simon India Limited
C
ontrol valves are one of
the most common and im-
portant instruments used
in the chemical process
industries (CPI). They help to en-
sure smooth and efficient operation
of process plants, by achieving the
desired operating parameters by
means of regulating the fluid flow in
connected pipes. The need to prop-
erly specify control valves during the
design phase of a plant cannot be
overemphasized.
The size of a control valve is de-
rived from a parameter called the
flow coefficient (Cv), which is defined
as volumetric flowrate (in gal/min)
of water through the valve at 60°F
when pressure drop across the valve
is 1 psi (Cv is calculated using the
formula given in the standard ISA-
75.01.01-2007). Process engineers
should take the following aspects
into consideration when specifying
control valves, to ensure that the
valves that are manufactured by the
vendors function according to the
requirements.
1. Controllability. While specifying
a control valve during the design
phase, the process engineer should
ensure that the valve’s controllability
must be good over the entire range
between minimum and maximum
flowrates. This can be done by es-
timating the maximum Cv and mini-
mum Cv that correspond to maxi-
mum flowrate and minimum flowrate,
respectively. In general, the control-
lability of a control valve is deemed
acceptable if its travel at maximum
flowrate does not exceed 90% of the
rated travel, and if travel at minimum
flowrate is in the range of 10–20%
of the rated travel. This means the
ratio of estimated maximum Cv to
estimated minimum Cv should pref-
erably not be more than 15. If the
ratio far exceeds this value, travel
at minimum flow may be less than
10% of the rated travel, or the travel
84
at maximum flow may be greater
than 90% of the rated travel — both
scenarios mean poor controllability
of the valve. In that case, pressure
drop across the control valve should
be increased so that the target ratio
can be lowered, as shown in Equa-
tion (1). For incompressible fluids,
the ratio of maximum Cv to minimum
Cv is given by Equation (1):
(1)
Figure 1 shows a typical control
valve circuit. The following notes add
further explanation:
i. The segments represented by
P1A and BP2 represent items con-
nected to the valve (such as pipes,
fittings, heat exchangers, flow el-
ements, and more), whereas the
segment represented by AB rep-
resents the control valve in the
complete circuit P1ABP2. The ar-
rows in this circuit represent flow
direction.
ii. ∆P denotes pressure drop.
iii. Indicated pressure drops are for
maximum flow.
Referring to Figure 1, if x is the
pressure drop across the control
valve circuit P1ABP2, and y is the
pressure drop across the control
valve AB for maximum flow, then the
pressure drop in the remaining part
of the circuit (consisting of the pipes,
fittings, heat exchangers, flow ele-
ments and more; as represented by
the segments P1A and BP2) is x–y
for maximum flow.
If r is the ratio of maximum to
minimum flow, and z is the pressure
drop across the circuit P1ABP2 at
minimum flow, then — ignoring the
elevation difference between P1 and
P2 — pressure drop in the part of the
circuit other than the control valve
(that is, P1A and BP2) at minimum
flow is approximately = (x–y)/r2.
Pressure drop across the con-
trol valve AB at minimum flow is
approximately:
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
P = y
P1
A B
P2
P = x
FIGURE 1. Shown here is a typical control valve
circuit, which is used illustratively in the descrip-
tions provided in the main text
(2)
From Equation (1) and Equation (2),
(Cv)max/(Cv)min is approximately:
(3)
The following conclusions can be
drawn from Equation (3):
iv. As r>1 and z≥x, any increase in
y leads to decrease in the ratio
(Cv)max/(Cv)min. That is, better
controllability can be achieved by
increasing pressure drop across
the control valve at maximum flow.
(Note: z = x if P1 and P2 are fixed
pressure points, and in general z>x
if the control valve is located at the
discharge of a centrifugal pump
v. If r increases, y also increases
for the same ratio of (Cv)max to
(Cv)min. This means that pressure
drop across the control valve at
maximum flow should increase
with an increase in the ratio of
maximum to minimum flow, if
the same controllability has to
be achieved.
vi. In most common cases,
maximum flow is 110%, and mini-
mum flow is 50%, of normal flow.
In such cases, if pressure drop
across the circuit P1ABP2 remains
same for maximum and minimum
flows, then Equation (3) becomes
the following:
MARCH 2017
 Recognize and honor the most noteworthy
chemical-engineering technology by
nominating it for the Kirkpatrick Chemical
Engineering Achievement Award

Winner is selected by a Board of Judges

comprised of current chairs of chemical-
engineering departments at accredited U.S. and
E.U. universities. It is this unbiased selection
process, combined with a more than 80-year
tradition that makes the Kirkpatrick Chemical
Engineering Achievement Award one of the most
prestigious honors that a CPI company can The aim of the Kirkpatrick Chemical Engineering
receive. The Kirkpatrick Chemical Engineering Achievement Award is to recognize and honor the
most noteworthy chemical-engineering technology
Achievement Award will be bestowed in
commercialized anywhere in the world during the
November at the Chem Show in New York. two years prior to a given award year.
Chemical Engineering magazine has awarded this
For the full list of winners and biennial prize continuously since 1933.
submission details, please visit:
chemengonline.com/kirkpatrick

DEADLINE FOR NOMINATIONS:

MARCH 15, 2017

chemengonline.com/kirkpatrick
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017 85
 (Cv)max/(Cv)min is approximately =
[1 + (3.84x/y)]0.5, which implies that
(Cv)max/(Cv)min is approximately
3.5, for an x/y value of 3.0. A (Cv)
max/(Cv)min value close to 3.5 cor-
responds to reasonably good con-
trollability. Hence, it can be said
that for good controllability, pres-
sure drop across the control valve
should be approximately one third
of total dynamic pressure drop
across the circuit at maximum flow,
if maximum and minimum flow are
110% and 50%, respectively, of
normal flow.
Although Equation (3) is true for
incompressible fluid, the above re-
sults in general are reasonably true
for compressible fluid, as well.
2. Cavitation. When fluid is flowing
through a control valve, the mini-
mum pressure occurs at the vena
contracta, and then pressure in-
creases along the path of flow until
the fluid reaches the outlet of the
control valve. The vena contracta is
the point in the flow path where the
flow area is minimum, the veloc-
ity is maximum and, hence, pres-
sure is minimum [1]. For liquids, if
the pressure at the vena contracta
is less than the vapor pressure of
the liquid, vapor bubbles will form.
Downstream of the vena contracta,
pressure recovery takes place, re-
sulting in higher pressure at the
valve outlet than at the vena con-
tracta. If pressure at the outlet of
the control valve exceeds the vapor
pressure, the vapor condenses
and bubbles collapse. As bubbles
collapse, it causes impact on the
valve body and creates noise. This
phenomenon is called cavitation.
Full cavitation occurs when pres-
sure drop across the control valve
is more than or equal to certain
minimum pressure drop (or critical
pressure drop), and the pressure
at the outlet of the control valve is
more than the vapor pressure of
the liquid. Thus, full cavitation oc-
curs if the following conditions are
met [2]:
P FL 2 ( P1 FF PV ) (4)
P = P1 P2
Where:
P1 = Absolute pressure at the inlet of
the control valve
86
TABLE 1: COMPARISON OF VARIOUS TYPES OF CONTROL VALVES [1 ]
Parameter Type of the control valve
Butterfly valve Ball valve Globe valve
Cost Low Low High
Pressure drop Low Low High
Flow capacity High High Low
Flow characteristic Fixed (approximately Fixed. For V-notch ball Any of the three (linear,
equal percentage) valve, flow characteristic equal percentage, quick
is equal percentage opening), depending on
plug or cage design
Liquid pressure recovery Low Low High
factor (FL)
Cavitation potential High (due to low FL) High (due to low FL) Low (due to high FL)
Noise potential High High Low. Can be reduced
further by using cage
guided plug
Control range Low In-between for V-notch High
ball valve
Control type Throttling and on-off On-off. However, V-notch Throttling
ball valves have good
throttling capability, too
Rangeability Low V-notch ball valves have High
very high rangeability
P2 = Absolute pressure at the outlet in an exchanger in the control
of the control valve valve circuit, the control valve
FL = Liquid pressure-recovery factor, should be located upstream of
as defined by: the exchanger — not down-
0.5 stream — so that vapor pressure
FL = [(P P2 ) /(P1 PVC )] of the liquid is sufficiently less
than the operating pressure at
Pvc = Absolute pressure at the vena the vena contracta of the control
contracta valve. This will eliminate flashing
FF = Liquid critical pressure ratio fac- through the control valve.
tor iii. Locate the control valve in the
FF = 0.96–0.28 (Pv / Pc)0.5 circuit where the elevation is mini-
Pc = Absolute thermodynamic criti- mum so that static head is maxi-
cal pressure of the liquid mum. This will maximize P1 as well
Pv = Absolute vapor pressure of the as P2 without impacting (P1–P2).
liquid at inlet temperature That can make the left-hand term
The exact value of FL for a particu- in Equation (4) less than the right-
lar valve can only be available in the hand term and, hence, prevent
valve vendor’s specifiction sheet, but cavitation.
an indicative value can be obtained iv.Select a control valve body type
from the vendor catalog, from ISA- with higher pressure-recovery
75.01.01-2007, or from other con- factor (FL), which makes critical
trol valve literature during the design pressure drop for cavitation —
phase. that is, FL2(P1—FF PV) — higher.
Process engineers should try to For example, a globe valve has
minimize the possibility of cavitation a higher FL than that of butterfly
when specifying control valves. This and ball valves. Thus, the use of
can be done in the following ways: a globe valve (instead of a butter-
i. Alter the hydraulics of the control fly or ball valve) might prevent the
valve circuit and reduce the pres- cavitation.
sure drop across the control valve It is absolutely necessary to specify
to less than FL2(P1—FFPV), where the thermodynamic critical pressure
possible (as explained above), and vapor pressure of the liquid in
without compromising the control- the process datasheet of a control
lability of the valve. valve handling a liquid, so that the
ii. Try to change the location of the occurrence of the cavitation can
control valve such that the possi- be assessed. Full cavitation re-
bility of flashing through the con- sults in choked flow and hap-
trol valve can be avoided. For ex- pens if ∆P ≥ FL2 (P1–FFPv). How-
ample, if a liquid is being heated ever, partial cavitation can occur
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM MARCH 2017
without causing choked flow, if
pressure drop is less than criti-
cal pressure drop but greater than
∆Pincipient cavitation [2],where:
∆Pincipient cavitation =
KC ( P1 FF PV )
Where:
Kc = The coefficient of incipient cavi-
tation, which is less than FL [2].
3. Multiple operating cases.
Whereas a control valve is gen-
erally specified for three cases
— minimum, normal and maxi-
mum flowrates, with correspond-
ing pressure drop — there may be
more than three operating cases.
In such situations, normal flowrate
and corresponding pressure drop
should be specified in accordance
with the normal operating case,
whereas other operating cases (if
there are more than two) should
be narrowed down to two cases.
When narrowing down the operat-
ing cases, Cv should be estimated
for each case. Then, minimum and
maximum flowrates (and corre-
sponding pressure drops) should
be specified in such a way that
they correspond to the minimum
and maximum Cv of the control
valve, and the Cv corresponding to
all other cases should lie between
minimum Cv and maximum Cv. As
actual Cv is not available when a
control valve is specified, the esti-
mated Cv should be used.
4. Selection of type of valve. But-
terfly valves, which are compact and
generally have a relatively low cost,
are often the first choice. However,
constraints may dictate otherwise.
For instance, if high pressure drop
across the valve is required, a globe
valve may be the better choice.
Because the resistance of a globe
valve is higher than that of a butterfly
valve, higher pressure drop can be
obtained across a globe valve with
reasonable size.
In liquid applications, high pres-
sure drop could lead to cavitation.
As the pressure-recovery factor of
globe valves tends to be higher than
that of other valve styles, cavitation
can often be avoided with the use of
globe valves.
In the case of gases, high pres-
sure drop could lead to choking
flow conditions, which can generate
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
excessive noise. Noise can be mini-
mized with a globe valve with the
use of cage-guided trim. However,
if the available pressure drop across
the valve is low, then a butterfly
valve may be the preferred choice.
Meanwhile, V-notch ball valves
can be preferred where high range-
ability is required. Standard, round-
ported ball valves are generally used
for on-off applications. Table 1 pro-
vides valve-selection guidelines.
5. Leakage class. The allowable
control valve seat leakage is spec-
ified in terms of ANSI/FCI 70-02-
2006 leakage class. This standard
recognizes six classes of allow-
able seat leakage (Class I, II, III, IV,
V and VI). Class I means highest
allowable leakage; Class VI means
least allowable leakage [3]. Gener-
ally, control valves for CPI applica-
tions are specified with leakage
Class IV. However, in situations
where tight shutoff is required, at
least Class V should be specified.
If a control valve is discharging to
a flare, or is controlling (on-off) fuel
flow to the burner of a fired heater
or furnace, it should be specified
with Class VI leakage.
6. Flow characteristics. The most
common types of inherent flow char-
acteristics are the following [1]:
• Linear — A valve with an ideal
linear inherent flow characteris-
tic produces a flowrate that is di-
rectly proportional to the amount
of valve plug travel, throughout the
travel range.
• Equal percentage — Ideally, for
equal increments of valve plug
travel, the change in flowrate re-
garding travel may be expressed
as a constant percent of the flow-
rate at the time of the change.
• Quick opening — A valve with
quick-opening flow characteris-
tic provides a maximum change
in flowrate at low travel rates. A
quick-opening characteristic is
basically linear through the first
40% of valve plug travel (corre-
sponding to 70% of maximum
flowrate), and there is little in-
crease in flowrate with further in-
crease in plug travel.
The flow characteristic of a valve de-
pends on its trim design. While ball
valves and butterfly valves have fixed
characteristics, globe valves can
have any of the three characteristics,
depending on plug or cage design.
MARCH 2017
The type of flow characteristic
should be specified in the process
datasheet, considering the param-
eter to be controlled, or the pres-
sure drop scenario in the system.
For flow or level control, linear char-
acteristic should generally be speci-
fied. In general, linear characteristic
should also be specified if most of
the pressure drop (as a proportion of
total pressure drop in the system) is
across the valve itself so that pres-
sure drop across the valve remains
nearly constant for varying flow-
rates. Equal-percentage character-
istic should be specified for pressure
control, or where a high proportion
of the total pressure drop occurs
in the system other than the valve
(that is, in pipes, fittings, equipment
and so on). It should also be speci-
fied where pressure drop across
the valve varies with varying flow-
rate. A quick-opening characteristic
should be specified for on-off appli-
cations. As in most of the common
systems, pressure drop across the
control valve varies significantly with
flowrate, so equal-percentage flow
characteristics are most commonly
specified [1]. Q
Edited by Suzanne Shelley
References
1. Emerson Process Management, “Control Valve Hand-
book,” 4th Ed., p. 18, pp. 33–36, p. 46, pp. 59–60, pp.
109–110, p. 136.
2. “Masoneilan Handbook for Control Valve Sizing,” 7th Ed.,
pp. 7–8, p. 10, 1987.
3. American National Standard, Control Valve Seat Leakage,
ANSI/FCI 70-2-2006, Cleveland, Fluid Controls Institute,
Inc., p. 2, 2006.
Author
Satyendra Kumar Singh is the
general manager (Head of Depart-
ment) – Process for Simon India
Limited (Mehtab House, A-36,
Ground Floor, Mohan Co-operative
Industrial Estate, New Delhi-
110044, India; Email: sty_singh@
yahoo.com; satyendra.singh@
adventz.simonindia.com). He has
more than 24 years of experience
in engineering consultancy and engineering procure-
ment construction in the fields of petroleum refining,
petrochemicals, chemicals, oil-and-gas, and ammonia.
Singh holds a Bachelor of Technology (Honors) degree
in chemical technology from Harcourt Butler Techno-
logical Institute (Kanpur, India), and an Masters of Busi-
ness Administration from Indira Gandhi National Open
University (New Delhi, India). He is a Chartered Engineer
(India), and a Fellow of The Institution of Engineers
(India); and has published numerous papers on man-
agement and engineering subjects.
87
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 6/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 7/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 8/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 9/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 10/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 11/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 12/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 13/15
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/29 14/15

Keys to successful
alarm management
Here are the features you want in alarm management software
J. GOOCH, Consultant, Houston, Texas
O
ver the past few years, alarm
management has become a very
important topic within a number
of articles, technical papers and books.
Most of the discussions have been centered
on the importance of alarm management
and how to implement and maintain an
alarm-management process. However, not
much has been written on the tools to sup-
port alarm management and, in particular,
alarm-management software. This article is
intended to meet that need.
A number of alarm-management soft-
ware systems are on the market. As you
might expect, each has its own strengths
and weaknesses. This article will provide a
vendor-neutral listing of key features, along
with a brief discussion of each.
Connectivity. Although alarms can be
generated on many systems such as dis-
tributed control system (DCS), emergency
shut-down (ESD), tank-gauging or flow-
measurement systems, most companies
bring alarms to the DCS since it is used as
the primary device for presenting informa-
tion to the operator. Therefore, it is abso-
lutely vital that the alarm-management
software has good connectivity to the DCS.
Numerous ways to connect an alarm-man-
agement system (ALMS) to a DCS include:
• Object linking and embedding for
process control (OPC) data access (DA)
• OPC alarms and events (AE)
• Printer intercept
• Open database connectivity (ODBC)
• Event files.
The method does not matter much as
long as the connectivity scheme is stable,
collects information on all alarms and can
collect at least a minimum set of parameters
for each alarm. A suggested minimum list
of parameters is:
• Alarm time/date
• Tag name
• Description
• Alarm [e.g., process variable high
(PVHI), process variable low (PVLO),
change of state, etc.]
• Value at time of alarm.
For certain advanced features, it is use-
ful if the connectivity solution provides the
ability to write information back to the
DCS. However, keep in mind that it is fun-
damental that the ALMS had good data
with which to work. Therefore, if the ALMS
you are considering does not have rock-solid
connectivity, find another ALMS. Period.
Alarm analysis. Another fundamental
feature is a suite of alarm analyses. Infor-
mation from alarm analysis will be used
to drive alarm monitoring and reduction
efforts, as well as to provide information for
reporting mechanisms. The bare minimum
analyses to be included are:
• Alarm frequency—A listing of alarms
in order of frequency. The analysis should
be able to show what percentage of total
alarms comes from each alarm reported.
• Alarm rates—A listing of alarm rates
per unit of time over an extended period,
for a selected operator position. Typical
time units are per day, hour and 10 mins.
The analysis must be able to distinguish
between annunciated alarms and recorded
alarms. This type of analysis can be used
to give an estimate of operator and system
alarm loading.
• Alarm floods—Alarm floods can be
the bane of an operator’s life. Alarm floods
are generally considered to occur when the
alarm rate exceeds 10 alarms per 10-mins.
It is commonly considered that an alarm
flood is a time when the system generates
more alarms than the operator can process.
It represents a time when the alarm system
is of no value.
• Chattering alarms—An alarm that
occurs three or more times in a minute is
SAFETY
considered to be chattering. A listing of chat-
tering alarms can be used by the maintenance
and engineering groups to target resources on
instrument and process problems.
Additional, very useful alarm analyses are:
• Duplicate alarms—A listing of alarms
where one alarm always follows another.
Effectively, the alarms are duplicates and
the removal of one should be considered.
• Stale alarms—A listing of alarms that
have been in the alarm state for more than
24 hrs. A stale alarm is not available to the
operator and may be considered disabled.
Such an alarm should be reviewed and
repaired or removed.
• Dynamic alarm priority distribu-
tion—A distribution of alarms by prior-
ity. This can be an indication of a problem
with alarm priorities.
• Operator changes—A listing of
changes initiated by an operator, such as
changes to alarm, controller state or con-
troller setpoints. A high change rate is
indicative of problems with the alarm sys-
tem or control loops. The more time an
operator spends adjusting the system, the
less time is spent managing the unit.
The alarm analysis system should sup-
port a very flexible set of alarm system
parameters. This is generally accomplished
by the use of filters that allow selecting such
things as:
• A specific operating area or set of
areas
• Time periods such as past seven days,
complete week, 30 days or complete month
• Alarm type
• Alarm priority.
If a site can get control of the most fre-
quent and chattering alarms and the major
alarms during alarm floods, the average
alarm rates will come down. Lowered alarm
rates during an abnormal situation give the
operator time to respond to the alarm. This
means the operator is thinking rather than
I
HYDROCARBON PROCESSING APRIL 2011 85
SAFETY

reacting. However, lowered alarm rates do of the best ways to determine an effective selection criteria are much too complex to
not solve the entire problem. The alarm alarm set for a control system. Typically, a be discussed here. Please see the references
system must be reviewed to ensure that the site will develop a formal document, often included at the end. The ability to docu-
optimum collection of alarms and priorities called an alarm philosophy, that details for ment the alarm rationalization results is
are included. alarm selection and priority determina- crucial. This will allow others to determine
tion criteria. The ALMS must support the the thought process later. Also, the alarm
Alarm rationalization support. alarm selection and priority determination system is essentially part of the unit system.
An alarm rationalization (also known as criteria. Most sites use a selection mecha- Being able to show how and why an alarm
an alarm objective analysis, or AOA) is the nism based on the maximum severity of has been set may be of importance during
systematic review of all potential alarms the consequences if there is no response a thorough review of such systems.
within an operating area, applying a rig- to the alarm and the time available for the Every alarmable parameter for each tag
orous set of selection criteria. This is one operator to react. The details of developing in the system must be presented for review.
This implies some knowledge of the sys-
tem configuration by the ALMS which,
in turn, implies that some mechanism for
importing configuration information into
Compression… the ALMS. This is not trivial. The vendor
must be able to demonstrate a reliable con-
figuration import method for all applicable
Without Exception control systems.
Other desirable features of an alarm
rationalization support package are:
• Filtering/selection capabilities—it is
often good practice to look at a group of tags
together. For example, it might be prudent
to look at all of the tags related to a distilla-
tion column to create a consistent column
alarm strategy for the column, as opposed to
looking at individual tags in isolation.
• Copy/paste capabilities—once an
alarm strategy is determined for one piece
of equipment, it may be possible to use it
for other similar pieces of equipment. This
has the advantage of enforcing consistent
Howden Burton Corblin Process Compressors strategies.
from Gas and Air Systems • Export of results—This can ease
developing management of change (MOC)
When you specify a compressor • Two-compartment, documentation.
to API 618, you expect the high full-stroke distance pieces. • Results implementation—Results
reliability achieved by compliance of an alarm rationalization are only useful
• Buffer-purged packing cases.
with the Standard. if actually implemented. There are often
• Conservative design hundreds, if not thousands, of changes that
Gas and Air Systems’ Howden
and ratings. must be made to the alarm system. ALMS
Burton Corblin Compressor
packages include the following support is convenient and a reliable method
API 618 demands it…you
important features required by to transfer the results to the control system.
expect it, and we provide it...
API 618...without exception. without exception.
Reporting. Reports often form the heart
• Water-cooled, horizontal Contact Gas and Air Systems of an alarm-system improvement process.
cylinders. for the process compressor built It has been said that if you cannot measure
• Replaceable dry cylinder liners. to your standard. an activity, you cannot improve it. Most
ALMS reporting packages are based on the
If you require low-volume and/or high-pressure compression analysis package. Reports should be:
of hazardous gas with zero leakage and zero contamination,
• Configurable to meet site needs
Gas and Air Systems’ Howden Burton Corblin Diaphragm Compressors
• Able to run automatically
will meet your specification…without exception.
• Available in multiple formats such as
HTML, spreadsheets such as XML or flat
Gas and Air Systems, Inc.
1304 Whitaker St. • Hellertown, PA 18055
files.
Toll free: 1 (888) 8 GASAIR • Ph: (610) 838 9625 Most companies separate the control sys-
Fax: (610) 838 9650 • info@gasair.net tem from the IT business LAN by a firewall
www.gasair.net of some type. Most ALMS packages will be
on the control-system side of the firewall.
Select 174 at www.HydrocarbonProcessing.com/RS
86
Select 78 at www.HydrocarbonProcessing.com/RS
AD: www.graficadueprint.com
© 2010 Costacurta S.p.A.-VICO

SINCE 1921...
AND WE
STILL LOVE IT

For more than eighty years, we at Costacurta have been

constantly and resolutely committed to the development
and manufacture of special steel wire and plate
components used in many different industrial processes.
Every day at Costacurta, we work to improve the quality
of our products and services and the safety of all our
collaborators, paying ever-greater attention to the
protection of the environment.
Within the wide range of Costacurta products you will
also find some, described below, that are used specifically
in the oil, petrochemical and chemical industries:
- RADIAL FLOW AND DOWN FLOW REACTOR INTERNALS;
- GAS-LIQUID AND LIQUID-LIQUID SEPARATORS;
- ARMOURING OF REFRACTORY, ANTI-ABRASIVE AND
ANTI-CORROSIVE LININGS.
For more information visit our website or contact the
division 'C' components for the oil, petrochemical
and chemical industries at tcrc@costacurta.it.
Radial Flow and Down Flow reactor internals

Costacurta S.p.A.-VICO Management systems

via Grazioli, 30
20161 Milano, Italy
tel. +39 02.66.20.20.66
fax: +39 02.66.20.20.99
www.costacurta.it
certified by LRQA:
ISO 9001:2008
ISO 14001:2004
OHSAS 18001:2007
Select 84 at www.HydrocarbonProcessing.com/RS

Reports are likely to be required on the busi-
ness LAN for ease of reporting to manage-
ment. Some method for transporting reports
across the firewall will be beneficial.
Many companies use key performance
indicators (KPIs) to measure performance.
KPIs for the alarm system will likely be
developed as a part of the alarm philosophy.
The reporting package should have some
means of generating reports that match the
KPI requirements; otherwise, the site will
need to process data to generate the required
KPI. For example, import XML files into
Excel and calculate the KPI data there.
Advanced features. The basic features
listed previously can be used to improve
alarm-system performance by focusing on
remedial activities, developing an optimal
set of alarms, and providing reporting facil-
ities to support ongoing improvement pro-
cesses. These features assume that the alarm
system is static. There are features that are
more real-time, where the alarm system is
modified based on input from the system,
or on predefined operator input. This is a
broad field so only a few concepts can be
briefly discussed.
• Audit and enforcement—Without
attention, alarm systems will degrade and
drift from the settings determined during
alarm rationalization and implemented
by MOC processes. Inappropriate change
comes from many sources and is, unfor-
tunately, common. An alarm audit is the
periodic comparison of the alarm-system
settings with the values implemented. Lists
of discrepancies can be reported as needed.
Also, the ALMS can correct discrepancies
by implementing the correct value in the
control system.
• Alarm shelving—Alarm-system mal-
functions will occur at the most incon-
venient times. Alarm shelving allows for
the temporary alarm suppression in a con-
trolled fashion. Care must be used with this
feature since an alarm that is shelved forever
is no alarm. To prevent this, alarm shelving
should have the following characteristics:
o Allow shelving of only one alarm
at a time, not all alarms for a tag.
o Provide security for access.
o Provide control over the maxi-
mum shelving time allowed.
o Record the identity of the person
shelving the alarm.
• State-based alarms— Since some
equipment can exist in multiple states (such
as out-of-service, half rate or alternative feed
stock), different alarms may apply at dif-
ferent times. Once the different states have
SAFETY
BORSIG
been identified, alarms for the affected tags
will be rationalized for each state. The state-
based alarm package will implement the
different alarms as the different states occur.
This can be triggered by a change in a tag
value, an operator action or a calculation. A
good state-based alarm package will incor-
porate a method of performing a calculation
to control the transitions between states.
• Alarm flood suppression—despite all
efforts to create an optimum alarm param-
eter set, alarm floods may still happen,
especially during an upset. An alarm-flood
suppression package can help in this situ-
ation by detecting an abnormal situation
and adjusting the alarm system by either
modifying alarm priorities/settings or by
suppressing the alarms. The flood trigger
must be detectable and the alarms to be
included must be known. The package
must also be able to reverse the changes
appropriately. It should be noted that it
takes time for the event to be detected and
the alarm changes made. There may well
be an initial alarm burst, but the alarm rate
will fall away very quickly, which reduces
demands for operator attention.
These packages must be aware of each
other and work together. For example,
the audit and enforcement package must
Leading
be aware of changes made by the alarm-
shelving package and the state-based alarm
Technology for
package to prevent errors in reporting and Innovative
potentially enforcing the wrong alarm val-
ues. And there you have it—a brief discus- Solutions
sion of features that, in my opinion, are in
an ALMS. Hopefully this will provide some Quench Coolers
food for thought if the need should arise to Process Gas Waste Heat
evaluate an ALMS. HP Recovery Systems
for ammonia, methanol, hydrogen, gas-to-liquid,
coal gasification and direct reduction plants in
BIBLIOGRAPHY
iron metallurgy, waste heat boilers downstream
The Equipment and Materials Users’ Association of the partial oxidation of oil and natural gas
Publication No. 191, Alarm Systems: A Guide to
Design, Management and Procurement, Scraped Surface Exchangers
www.eemua.org. Reciprocating Compressors
ANSI/ISA—18.2-2009, Management of Alarm for Process Gases
Systems for the Process Industries, www.isa.org.
Hollifield and Habibi, The Alarm Management Centrifugal Compressors
Handbook, www.pas.com. for Process Gases
Compressor Valves
Jerry Gooch has over 30 years of
experience in the oil and gas industry. Other BORSIG Products:
The majority of this experience has Membrane Technology, Boiler and Power Plant
been in process control in refineries Technology, Industrial and Power Plant Services
in the US and the Middle East with a
notable exception as the controls coordinator for the new
build liquefied natural gas ships in Japan. Mr. Gooch was BORSIG GROUP
employed as an alarm management consultant for several
Egellsstrasse 21
years, executing alarm management projects at refining,
D-13507 Berlin/Germany
petrochemical and power plants in the US, Europe and
the Middle East. He worked closely with plant managers,
Phone: +49 (30) 4301-01
Fax: +49 (30) 4301-2236
engineers and operators and was a strong proponent of
a comprehensive operator effectiveness program’s value E-mail: info@borsig.de
and alarm management’s key role in this program. Mr. www.borsig.de
Gooch is currently employed at KBR in Houston.
Select 175 at www.HydrocarbonProcessing.com/RS
LOOP CONTROL

Best practices for PID

by Gregory K. McMillan and P. Hunter Vegas

PROPORTIONAL-INTEGRAL-DERIVATIVE (PID) control is used in more than 99% of the

loops for chemical, pharmaceutical, and food and beverage production. PID has proven
to be capable of the best load rejection by using the best tuning and the extensive flex-
ibility of functionality offered by the PID structure and options, most notably external-
reset feedback. Bohl and McAvoy enlightened us about it in a 1976 landmark paper,
“Linear Feedback vs. Time Optimal Control, II. The Regulator Problem” (Industrial & En-
gineering Chemistry, Process Design and Development, Vol. 15, No. 1, 1976, p. 30-33).
The opening challenge with PID control is recognizing the effect of PID form, struc-
ture and tuning units. Going from proportional band in percent to a dimensionless PID
gain, and from a reset in repeats per minute to seconds, are recognizable factors to
address, but are sometimes overlooked. Less understood is how to convert tuning set-
tings to the ISA Standard Form from the Series (Real) Form in analog control and early
DCS systems, and from the Parallel (Independent) Form in the control literature, con-
verting integral and derivative gains to reset and rate times.

PID BASICS
1. Measure each flow, inventory (e.g., level or weight and pressure), every process variable for safe
operation and environmental protection, and each indicator of process efficiency and capacity
(e.g., composition, quality, pH, and temperature) with sufficient accuracy (particularly least drift
and best repeatability), speed and rangeability.
2. Manipulate each flow by a control valve or variable-frequency drive (VFD) with the least dead-
band (e.g., < 0.4% backlash) and sufficient linearity (e.g., < 5:1 gain change), resolution (e.g.,
< 0.2% stiction), speed (e.g., < 2-sec., 86% response time) and rangeability (e.g., > 40:1).
3. Make sure every component in the process, whether a reactant, inert, byproduct or product, is
accounted for by field or scheduled lab measurements, and is controlled to ensure exit or con-
sumption by reaction with no continual accumulation.
4. Pair controlled variables (e.g., composition and temperature) with manipulated variables (e.g.,
flows) that have the largest effect as seen in Relative Gain Array (RGA).
5. If the RGA assessment of choices doesn't clearly favor a particular pairing, choose the manipu-
lated and controlled variable pair with the smallest deadtime-to-time constant ratio.
6. Use flow feedforward for gas pressure control with linearization of valve characteristics.
7. Use flow ratio control for composition, level and pH control, where there's a secondary flow con-
troller and a ratio controller (RC) with a setpoint and bias that both can both be remotely set or
manually adjusted by the operator. If the RC block doesn't have a remotely settable bias, the RC
output becomes the input to a bias and gain block.
continued on p.26

22 • DECEMBER 2018 www.controlglobal.com DECEMBER 2018 • 23

Measurably drive
Measurably drive

Safely
Safely

Turn industrial automation into the profit engine of your business.

Turn industrial automation into the profit engine of your business.

A
A control
control system
system that
that measurably
measurably improves
improves operational
operational profitability?
profitability? Now
Now that
that is
is aa bold
bold idea.
idea.
Only the
Only the EcoStruxure
EcoStruxure Foxboro
Foxboro DCS
DCS provides
provides measurable
measurable operational
operational profitability
profitability improvements
improvements
and a
and a future-proof
future-proof architecture
architecture to
to enable
enable 100%
100% ROI
ROI in
in less
less than
than one
one year.
year. Together
Together with
with our
our
IIoT-ready apps, analytics,and services; edge control; and connected products – Schneider
Electric can optimize every asset in your enterprise and empower you to control operational
efficiency, reliability, safety, and profitability together – in real time.

#WhatsYourBoldIdea
#WhatsYourBoldIdea

schneider-electric.com/foxboro
schneider-electric.com/foxboro

Ignite your profit engine today by downloading our whitepaper.

Ignite your profit engine today by downloading our whitepaper.

© 2018 Schneider Electric. All Rights Reserved. Schneider Electric | Life Is On is a trademark
and the property of Schneider Electric SE, its subsidiaries, and affiliated companies. • 998-20278420_GMA-US
© 2018 Schneider Electric. All Rights Reserved. Schneider Electric | Life Is On is a trademark
and the property of Schneider Electric SE, its subsidiaries, and affiliated companies. • 998-20278420_GMA-US
 Your Global Automation Partner
Delay Lag Gain
θd τL Кd
∆DV
Load upset
Secondary Secondary Primary Primary
Delay Lag Gain delay lag delay lag
θV τV КV θS τS θP τP КP
∆FV
Valve or VFD Process ∆PV

Local
% ∆%CO setpoint Кm = 100%/span
%SP
PID КC Ti Td ½ of wireless default update rate
% ∆%PV (∆CV)
Delay Lag Gain Lag Delay Lag
τC2 θC τC1 Кm τm2 θm2 τm1 θm1
Lag
Controller Measurement Delay

MAJOR SOURCES OF LOOP DYNAMICS

Figure 1: For a first-order approximation, the total loop deadtime θ0 ≈ θV + θS + θP + θm1 +
θm2 + θC + Y*τV + Y*τS + Y*τm1 + Y*τm2 + Y*τC1 + Y*τC2. The Y fraction of small lag that is
equivalent to deadtime is a logarithmic function of the ratio of the smallest to largest lag (Y =
0.28, 0.88 for ratios of 1.0 and 0.01, respectively). КV is the slope of the installed flow charac-
teristic, and τ0 is the largest lag in the loop, hopefully τP. For self-regulating processes, К0 =
КV * Кd * Кm; for near-integrating processes, Кi = КV * (КP/τP) * Кm.

Especially noteworthy is that in the Series
Form, the derivative time (rate time) could be
set as large as the integral time (reset time).
Some older tuning methods even suggested
this. Interaction between these modes in
the time domain by Series Form would in-
herently prevent oscillations by effectively
reducing integral and proportional action
thru interaction factors. When using the ISA
Standard Form, the rate time has to be set
less than one-fourth the reset time to prevent
oscillations. Many users have made mistakes
in system upgrades by not paying attention
to the changes in form and units by the even
the same system supplier.
Literature can’t always
be taken literally
The literature is also misleading by nearly
always citing a process deadtime, process
time constant and process gain when the ac-
tual dynamics depend on what's in the auto-
mation system (Figure 1). PID tuning needs
β Gain
to use the open-loop gain, open-loop time
constant (primary time
+ constant), secondary
* total∆loop deadtime.
time constant and * The
-
%SP
www.controlglobal.com
For reverse action,
error = %SP - %PV
open-loop gain is the product of the valve or Reduce Costs – One
variable-frequency drive (VFD) gain, process
Sensor Detects
gain and measurement gain. The result-
ing gain is dimensionless for self-regulating all Metals
processes and has units of 1/sec for integrat-
ing processes. The open-loop time constant Detect all metals at the same
is the largest time constant in the loop, and range with inductive sensors
hopefully in the process to slow down distur- from Turck – no position
bances on the process input. Unfortunately adjustment required. These
for flow, liquid pressure and surge control, non-contact, durable inductive
the largest time constant is in the automation sensors help you reduce the
system, slowing down the ability to see and number of sensors needed
correct for disturbances. for detecting the position or
The secondary time constant is the sec-
presence of metal parts. Plus,
ond largest time constant in the system,
our IP68-rated versions stand
often in either heat transfer surfaces or the
up to challenging washdown
automation system. The total loop deadtime
environments.
is the sum of all the deadtimes and the frac-
tion of all the small time constants creating
equivalent deadtime. To better communi-
cate, realize the implications of dynamics,
and achieve the best PID, see the Control
Talk blog post, “Understanding terminol-
ogy to advance yourself and the automation
profession” (www.controlglobal.com/blogs/
*P = (β – 1) * Gain * %SP
controltalkblog/understanding-terminology- 1-800-544-7769
www.turck.us
For zero error,
Out1 = 0
LOOP CONTROL

to-advance-yourself-and-the-automation-
PID BASICS, continued from p.22
profession)
8. Ratio control (RC) details: Another major point of confusion is lack of
a. Connect BKCAL_OUT from a manipulated RC setpoint or bias to BKCAL_IN recognition that nearly all disturbances are on
of the PID, correcting the RC to ensure a bumpless transfer to the remote the process input and not the process output,
setting of the RC setpoint or bias. as is commonly depicted in the control litera-
b. The RC input is the leader flow (e.g., largest feed). The RC output is the cas- ture. Also, except for flow and liquid pressure
cade setpoint of the secondary flow controller, or is the input to a bias and control, the response is near-integrating or true
gain station whose output becomes the cascade setpoint of the secondary integrating, requiring integrating process tun-
flow controller. ing rules and an overshoot by PID output of its
c. F or vessels and columns, the primary controller corrects the RC bias. final resting value that's the output after com-
d. For inline plug flow systems, the primary controller corrects the RC setpoint. pletion of the response to a load or setpoint
e. For reactants, there may be several flow loops ratioed to the leader flow (e.g., change. What's in the literature are self-regulat-
primary reactant flow). To eliminate temporary stoichiometric imbalances ing processes with a process time constant-to-
from changes in production rate, put a setpoint filter with a filter time on each deadtime ratio for dynamics that rarely gets be-
flow loop just large enough for a smooth response, so reactant feed flows low 0.25 or above 4, leading to Internal Model
move in concert. Control. While these dynamics are infrequent
9. For override control of constraints (e.g., maximum pressure, temperature, or in chemical plants except for flow loops, they're
throttle valve position), use external-reset feedback to intelligently prevent applicable to pulp and paper and mining plants
unwanted integral action; tune each PID gain to provide the correct point of due to a lack of back mixing.
takeover (selection) of the override PID relative to its setpoint; historize the You can convert back and forth between
total time each PID is selected; and capture the overshoot of the setpoint. self-regulating and integrating process dynam-
10. Use cascade control to: ics by realizing that open-loop, near integrating
a. Provide rapid correction of secondary upsets and nonlinearities (e.g., secondary process gain is simply the open-loop self-reg-
flow loop correction for pressure changes and installed flow characteristic). ulating process gain divided by the open-loop
b. Enable flow ratio control by manipulating a flow PID setpoint instead of a valve. time constant (Figure 1).
c. Give a relatively constant and easily estimated open-loop process gain from Integrating process tuning rules are de-
controller output (CO) scale needed to prevent violating the low PID gain limit for signed to provide the best disturbance re-
processes with a near-integrating, true integrating or runaway response. jection. This can be tested by momentarily
d. Convert the secondary time constant from a detrimental term in the pri- putting the PID in manual and making the
mary loop creating deadtime to a beneficial term as the largest time con- largest step change in PID output that the
stant in the secondary loop, thus decreasing the deadtime-to-time constant operator is comfortable with (e.g., 5-10%).
ratio of both loops, and offering tighter control. Changes less than 1% typically show a much
e. Reduce the deadtime and hence the ultimate period of the primary PID, slower and possibly no response due to ac-
enabling better handling of disturbances originating in the primary loop by tuator and positioner sensitivity limit, as well
more aggressive tuning. as valve backlash and stiction.
11. Match the PID output limits and engineering units to the setpoint limits and
engineering units of the secondary loop. Set any anti-reset windup (ARW) Feedforward and ratio control
limits to match the output limits, except for special cases where you set the are underused
ARW limits inside the output limits when the PID must move quickly through Nearly all control loops end up manipulat-
stiction at a low limit, or quickly recover from a high limit that's an adverse ing a flow either directly by cascade control,
operating condition. where a primary process controller manipu-
12. Use valve position control (VPC) to maximize feed rate, waste fuel and waste lates a secondary controller setpoint, or by
reagent use, and minimize energy usage by minimizing compressor pressure directly positioning a control valve. Here,
and maximizing the refrigeration unit or cooling tower temperature by forcing we consider changes in speed as effectively
affected control valves (e.g., feed, waster fuel or reagent, coolant, steam and causing changes in flow even in sheet lines.
vent valves) to the maximum throttle position. Nearly all fast disturbances are changes in
13. Use positive-feedback implementation of integral action (Figure 2) for the ISA flow, and decoupling signals from interacting
Standard Form to provide true external-reset feedback (e.g., dynamic reset loops are changes in manipulated flow. The
limit). needed end result is a ratio of the manipu-
continued on p.28 lated flow or speed to a leader flow or speed
that is a disturbance from an uncontrolled

26 • DECEMBER 2018 www.controlglobal.com

 Your Global Automation Partner

MORE RESOURCES
To really dig into what PID can do, get the book, Tuning and Control Loop Performance,
Fourth Edition (www.momentumpress.net/books/tuning-and-control-loop-performance-
fourth-edition). See the following for a greater appreciation of what you should know
about PID control:
• www.controlglobal.com/articles/2017/the-greatest-source-of-process-control-knowledge
• www.controlglobal.com/articles/2017/common-automation-myths-debunked
• www.controlglobal.com/blogs/controltalkblog/solutions-to-prevent-harmful-feedforwards
• www.controlglobal.com/blogs/controltalkblog/uncommon-knowledge-for-achieving-
best-feedforward-and-ratio-control
• www.controlglobal.com/blogs/controltalkblog/deadtime-dominance-sources-conse-
quences-and-solutions
• www.controlglobal.com/blogs/controltalkblog/pid-options-and-solutions-part-1
• www.controlglobal.com/blogs/controltalkblog/pid-options-and-solutions-parts-2-and-3

(wild) flow, a manipulated flow by another
loop, or a deliberate change in production
rate or composition.
If there's no secondary controller, flow
feedforward is used primarily via a feedfor-
ward summer. The most common example
is steam header letdown pressure control,
where feedforward signals from steam users
and suppliers and decoupling signals from
other header letdown valves are summed
together in a feedforward summer. Signal
characterization is used where there are
not flow measurements to convert a per-
cent valve signal to a percent valve flow per
installed flow characteristics. For compres-
sor surge and pressure control, feedforward
summers are used for decoupling between
headers and disturbances from downstream
users, since the controller outputs go directly
to valves and vanes. Secondary flow control-
lers would be too slow in these cases.
Most process equipment involving liquids
can benefit from secondary flow controllers.
Here, ratio control is used to achieve and
correct as needed the ratios of flows seen
on a process flow diagram (PFD). Many pro-
www.controlglobal.com
cesses, particularly distillation columns, are
started up on ratio control. The ratio is cor-
rected by a bias to the secondary controller
setpoint for volumes with some mixing, such
as crystallizers, columns, evaporators, neu-
tralizers and reactors. For sheet lines and
extruders, the ratio setpoint is corrected.
Dynamic compensation must be applied
to a leader flow before it becomes a feed-
forward signal or ratio controller input flow.
To make sure a change in manipulated flow
arrives in the process at the same place at
the same time as a change in leader flow, a
deadtime and lead/lag (L/L) block is used.
If the correction arrives too soon or too late,
inverse response and sequel disturbance
are created, respectively. Unfortunately, tun-
ing software packages may not identify the
deadtime and L/L settings. The block dead-
time for dynamic compensation is the dead-
time in the disturbance variable path minus
the deadtime in the manipulated variable
path. The L/L block lead time is the lag time
in the manipulated variable path, and the
L/L lag time is the lag time in the disturbance
path. The lag time should be at least one-
Reduce PLC Costs
with Multi-protocol
On-Machine I/O
Expand your communication
capabilities and reduce costs
with innovative multi-protocol
on-machine I/O. Whether you
are using one communication
protocol or several, Turck
ensures you get flexible,
future-proof fieldbus solutions.
Our vast range of rugged on-
machine I/O streamlines your
processes and is ideal for out of
the cabinet applications.
1-800-544-7769
www.turck.us
DECEMBER 2018 • 27
LOOP CONTROL

fourth the lead time to prevent noise. Model

PID BASICS, continued from p.26 predictive control (MPC) does this identifica-
14. U se external-reset feedback with setpoint rate limits to provide directional move sup- tion and correction for the user, which, along
pression to enable a VPC to provide a gradual, smooth optimization and fast getaway for with ability to avoid tuning override control-
abnormal conditions, and to prevent unnecessary crossings of the split-range point. lers, are the principal reasons why MPC
15. Use external-reset feedback to stop oscillations from deadband and stiction, a slow replaces PID for even a small applications
secondary loop, or slow control valve, and to provide enhanced PID action to prevent where a PID could suffice (e.g., one decou-
oscillations for increases in analyzer cycle time. pler, one feedforward and one constraint).
16. Use signal characterization of the controlled variable for pH titration curves and of the After good load disturbance rejection is
manipulated variable for installed flow characteristics, where the changes in process achieved, good setpoint response is possible
gain or valve gain, respectively, from changes in slope are greater than a factor of five. by using a filter lag equal to the reset time, or
17. Minimize total loop deadtime by faster and more precise measurements and valves, a structure with proportional and derivative
shorter transportation delays, faster fluid velocity and better mixing. action on the process variable and Integral
18. M inimize process noise by a sensor location that's less affected by cross-sectional action on error (PD on PV and I on E). If you
changes in velocity, phases, temperature and concentration. want a faster approach to setpoint, you can
19. Use middle-signal selection to prevent adverse reactions to a single sensor failure use a L/L instead of a filter on the setpoint
of any type, decrease reaction to noise or a slow sensor, reduce unnecessary main- with the L/L lag time equal to reset time, and
tenance, increase reliability and accuracy, improve diagnostics minimize lifecycle the L/L lead time equal to one fourth the lag
costs, and maximize onstream time and product quality. time, or a two-degrees-of-freedom (2DOF)
20. Choose loop objectives and choose the PID structure in the ISA Standard Form by structure with the setpoint weights for propor-
turning off modes and by setting setpoint weights and tuning the PID accordingly. tional and derivative action set equal to 0.5
Since there's is no interaction in the time domain like there is in the Series Form to and 0.25, respectively. Most of the literature
prevent derivative action from exceeding integral action, the rate time must not ex- concentrates on setpoint response, ignoring
ceed one-fourth the reset time. load response and advantages of lead/lag
21. Employ procedure automation and state-based control to utilize the best operator and and structure options.
process knowledge to eliminate any operator manual actions during startup, transi-
tions and recovery from abnormal situations, such as equipment and instrument fail- Remember external-reset feedback
ures or unusually bad disturbances. The other thing most users are not aware
22. Compute a future value of PV by passing the PV through a deadtime block and adding of is the incredible capability offered by
to the block input the change in PV during the block deadtime, that is, the block in- external-reset feedback (e.g., dynamic re-
put minus the block output. Block deadtime is greater than loop deadtime for a good set limit) to:
signal-to-noise ratio. Use the future PV value for a faster startup and batch cycle • Suppress oscillations from a slow sec-
time, where the manipulated flow is initially maximized and then set to its final rest- ondary loop, slow control valve or VFD,
ing value when the future value is projected to reach a setpoint or batch endpoint. and deadband and resolution limits.
23. To prevent snowballing buildup of recovered and recycled component concentration, • Provide directional move suppression
ensure there's a flow controller with a fixed setpoint for a given production rate some- to prevent unnecessary crossings of
where in the flow path from the discharge back as a feed input of the affected equip- split-range points; enable gradual opti-
ment (e.g., centrifuge, column, crystallizer, evaporator, reactor). mization and fast getaway by a VPC, a
24. Performance metrics should be put online and historized for key unit operations and fast-opening and slow-closing vent, or
the production unit. The ratio of total raw material or utility used to total product pro- recycle valve for surge control; and mini-
duced over a representative time (e.g., last 12 hours or last batch) gives a measure mization of upsets to other loops.
of process efficiency. The total product produced over this time gives a measure of • Achieve deadtime compensation as
process capacity. The total value of product in dollars in this time frame minus the good as a Smith Predictor or MPC with-
total cost of raw materials (e.g., reactants, reagents, additives and feeds from other out having to know the open-loop gain
units), utilities (e.g., steam, coolant and electricity) and fixed costs (e.g., staff and or time constant, or needing to provide
depreciation) is the profit in dollars. Most important is the trend and not the accuracy a special operator interface. Deadtime
of prices and costs. compensation is achieved by simply in-
25. Use a virtual plant (digital twin) for flexible and fast exploring, discovering, prototyp- serting a deadtime block in the external-
ing, testing, justifying, deploying, testing, training, commissioning, maintaining, trou- reset feedback path and tuning the PID
bleshooting, auditing and continuous improvement, showing the “before” and “after” much more aggressively (e.g., reset time
benefits of solutions from online metrics. can be decreased by an order of magni-
tude if deadtime is accurately updated).

28 • DECEMBER 2018 www.controlglobal.com

 Your Global Automation Partner

β Gain
+ *P = (β – 1) * Gain * %SP
* ∆ *
-

%SP
For reverse action, For zero error,
error = %SP - %PV Out1 = 0
+ P Out1 PI
Filter ∆ * Σ
Filter time =
Σ %CO
- Positive I Out2 reset time D
feedback Bias
*P
γ Rate Switch position
time Filter ∆ for external-reset
feedback
+
* ∆ * * Filter Derivative
- ERF
%PV Filter Filter time =
α * rate time

ISA STANDARD FORM WITH TRUE EXTERNAL-RESET FEEDBACK

Figure 2: All signals are percent-of-scale in the PID algorithm but inputs and outputs are in
engineering units. For structures with no P action, gain is zero for the proportional mode. Gain
is one for the integral mode, and equal to the PID block gain setting for derivative mode. Bias
is used as input to the reset time filter block when there is no integral action. Bias is the PID
output when the error is zero, and is filtered by the reset time whose best setting is reduced to
be about the deadtime. External-reset feedback (ERF) is, for example, secondary loop process
variable (%PVs) or fast valve readback.

Figure 2 shows the only known time
domain block diagram that details the ISA
Standard Form with the ability to achieve
eight different structures and true external-
reset feedback with the positive feedback
implementation of integral action.
Relative gain analysis is a powerful tech-
nique for accessing the type and degree of
interaction. The relative gain for a given loop
is its open-loop steady state gain with the
other loops open divided by its open-loop
steady state gain with the other loops closed.
The given loop is open for the tests with the
other loops open and then closed, to show
response without and with interaction, re-
spectively. A loop is considered open if the
mode is manual or remote output.
For integrating processes, the process
variable is translated to a rate of change,
so a steady ramp rate gives a steady state.
The relative gain is dimensionless, and thus
doesn't depend on the engineering units of
the loops involved. The relative gain doesn't
change when a flow ratio rather than a flow
is manipulated. Operating point nonlineari-
ties will affect the analysis but not changes
in time constants and deadtimes. Ideally,
pairing of loops should have a relative gain
www.controlglobal.com DECEMBER 2018 • 29
close to one or slightly higher. Negative
relative gains are disastrous. Often, the rela-
tive gain array can be simplified to show Industrial Ethernet
the relative gains for the four pairs possible Cables – Your Path to
between two controlled variables and two
manipulated variables (2 x 2). Smart Manufacturing
If you want a comprehensive, concise
Smart machines need higher
view of how to achieve the best PID con-
bandwidth to process the
trol, and gain an excellent understanding of
real-time data that optimizes
everything else needed for a successful au-
your manufacturing processes.
tomation project, look for the McGraw-Hill
Turck’s industrial Ethernet
handbook Process/Industrial Instruments
cables meet those needs with
and Controls, Sixth Edition, due out early
rugged, flexible, and multi-
in 2019. The handbook covers the design,
protocol solutions. With many
installation, calibration, commissioning and
options for customization and
maintenance of every part of an automa-
short lead times, Turck is your
tion system, concluding with a list of best
ideal partner for more cost-
practices for each of its 100 sections.
effective networks.
Meanwhile, the sidebar provides best
practices excerpted from the handbook’s
PID Basics section.
Gregory K. McMillan, Control columnist and
Hall of Fame member, can be reached at Greg.
McMillan@emerson.com, and P. Hunter Ve-
gas, engineering manager, Wunderlich-Malec
Engineering, can be reached at hunter.vegas@
wmeng.com. 1-800-544-7769
www.turck.us
 Focus
Temperature measurement
United Electric Controls
HART-compatible transmitter
eases temperature monitoring
Industrial sensors monitor tempera-
ture or pressure abnormalities by
comparing the process conditions
with pre-programmed setpoints in
a PLC or DCS controller. The HART-
compatible One Series 1XTXSW
transmitter line (photo) is said to be
the first to integrate 4–20-mA out-
put and two solid-state program-
mable relays (compatible with HART
7) for monitoring temperature or
pressure in safety, alarm and emer-
gency-shutdown applications. Typi-
cal electromechanical switches are
blind and cannot alert workers that
something is amiss with the device,
says the company. With this device,
however, users deploying a HART
communications networks can add
switches to the field assets used in
their predictive-maintenance strat-
egy and reduce routine inspections
and unplanned shutdowns. Combin-
ing HART and smart diagnostics in
one device, workers can manage all
of their instrumentation assets, in-
cluding switches, using a centralized
asset-management solution. A large,
back-lit display provides process val-
ues and setpoint programming. The
One Series 1XTXSW hybrid transmit-
ter switch effectively replaces a gage,
switch and transmitter, requiring only
one connection to the process. This
helps to improve reliability while re-
ducing lifecycle costs, and allows for
health and switch-status notifications
at about half the cost of a traditional
process-control transmitter, says the
manufacturer. The new One Series iX
line has worldwide hazardous loca-
tion approvals for Class I, Divisions 1
Omega Engineering and 2 (Zones 0, 1 and 2). — United
Electric Controls, Watertown, Mass.
www.ueonline.com
PID control simulator supports
temperature tracking
The Platinum Series controller dem-
onstration and training kit provides a
PID control simulator (photo) that is a
fully integrated temperature-control
system. It is suitable for use in both
Note: For more information, circle the 3-digit number on p. 66, or use the website designation.
26 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
closed-loop (PID) and simple on-off
temperature control, in both heating
and cooling applications. This unit
has a thermoelectric heating and
cooling element that is capable of
5–50°C setpoint, and a Type K ther-
mocouple for temperature readings.
The unit comes with panel-mounted
USB and Ethernet connectivity, an
alarm indicator, a 4–20-mA remote
setpoint potentiometer, a push-
button digital input, and more. Its
32-bit, 120-MHz ARM processor
provides full PID control process
with auto-tuning and a fuzzy-logic-
based, adaptive control algorithm.
According to the company, the unit
automatically adjusts and optimizes
the control-loop parameters based
on external environmental or control-
system deviations. A comprehensive
set of alarm functions may be used
to trigger output signals, or modify
the display colors for conditions
that are above or below a thresh-
old value, or in or out of the accept-
able range. Setup is easy, says the
company, thanks to an integrated
four-button keypad and a structured
menu system that eases naviga-
tion to only those parameters appli-
cable to the selected configuration
step. — Omega Engineering, Inc.,
Stamford, Conn.
www.omega.com
Measure temperature reliably
in high-voltage systems
To ensure optimal operational safety
for power plants and grids, critical
factors, such as hot spots inside
high-voltage transformers, power-
factor-correction capacitors and al-
ternators are monitored. Similarly,
temperature is routinely measured on
high-voltage distribution power lines,
and in gas-isolated power breakers
and power-factor-adjusted capaci-
tors. The FotempTrafo is a fiber-op-
tic thermometer that is available with
one to eight measurement channels.
Its compact design and height-ad-
justable DIN-rail clip allow for easy
installation in a cabinet. A wide selec-
tion of interfaces (RS-232, RS-485,
JANUARY 2016

Profibus and Modbus) makes com-
munication easy. A fast measurement
speed of 250 ms per channel allows
the direct detection of overloads. Its
measurement range is from –200 to
300°C, with an accuracy of ±0.2K,
says the company, which notes that
its use of gallium arsenide (GaAs) in
the sensor tip eliminates the need for
recalibration throughout the product
life. The probe sensor is completely
non-conductive, and the fiber-optic
sensors offer complete immunity
against radio-frequency interference
(RFI), electromagnetic interference
(EMI), nuclear magnetic resonance
(NMR) and microwave radiation—
Optocon AG, Dresden, Germany
www.optocon.de
Compact thermal imager
provides sharp images
The M12 7.8KP Thermal Imager
(photo) is said to have the highest
pixel density in its class, and its ex-
clusive Dual Sense pixel technology
provides sharp definition of hot and
cold details within any given image.
It is designed to support trouble-
shooting and predictive maintenance
activities. It is particularly useful for
determining thermal conditions asso-
ciated with hard-to-access systems.
Its Dual Sense technology allows the
imager to individually optimize pixels,
which helps to reduce or eliminate
the blurring of hot and cold details in
complex scenes, says the manufac-
turer. When the device is integrated
with the company’s Thermal Imager
Report Software, users can analyze
images and generate reports quickly
and easily. Users can download im-
ages from the thermal imager with
its included 8-GB SD card or micro-
USB connection port. Every image
capture also generates an image with
the .png file format, allowing images
to be emailed for convenience. —
Milwaukee Tool, Milwaukee, Wisc.
www.milwaukeetool.com
This temperature transmitter is
accurate and EMC-resistant
The Model T15 is a new digital trans-
mitter (photo) with analog output. De-
signed to process signals from resis-
tance sensors and potentiometers, it
is ready for operation in less than three
seconds, according to the company,
and offers a high basic accuracy of
0.1% of span, at a measuring rate of
CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
Milwaukee Tool
up to 20 measured values per sec-
ond. At the same time, the T15 fulfills
high standards of safety, for example,
in accordance with NAMUR recom-
mendations for EMC protection, sig-
nalling and sensor-break monitoring.
It is said to be the first transmitter
to be certified to the new EMC pro-
tection standard, DIN EN 61326-2-
3:2013. The T15 also has ATEX and
IECEx approvals for use in hazardous
areas. When used with the WIKA-
soft-TT software, the transmitter can
be programmed intuitively. Both
head-mounted and rail-mounted ver-
sions are available. — WIKA Alexan-
der Wiegand SE & Co. KG, Klingen-
berg, Germany
www.wika.com
Distributed temperature sensor
keeps a watchful eye
The DTSX 3000 Distributed Tem-
perature Sensor is a single device
that is capable of detecting minute
temperature changes, and monitor- WIKA Alexander Wiegand SE
ing the temperature distribution over
a long distance or across a wide
area. It can measure the temperature
along fiber-optic cables up to 50 km
in length (eight times the distance
possible with the company’s previ-
ous model), with a spatial resolution
of one meter. It is well suited for plant
and infrastructure maintenance ap-
plications that require the monitoring
of temperatures over long distances
or wide areas. Specific applications
include temperature monitoring of
the outer walls of high-temperature
industrial furnaces, the detection
of gas and liquid leaks in tanks and
other large production facilities, and
the detection of abnormal heat levels
in power lines and natural-gas-recov-
ery processes. — Yokogawa Corp. of
America, Newnan, Ga.
www.yokogawa.com
Thermocouple wire resists
oxidation at high temperatures
This company’s Type N thermocou-
ple wire provides excellent repeat-
ability and resistance to oxidation
at high temperatures (1,500°F and
above) and in applications where sul-
fur is present. The nickel-based ther-
mocouple alloy resists oxidation at
higher temperatures, and is designed
for applications with temperatures
from 200 to 2,000°F, such as high-
temperature heat-treating applica-
JANUARY 2016 27
 Quintex GmbH
tions. — Watlow Electric Manufac-
turing Co., St. Louis, Mo.
www.watlow.com
This electronic thermostat has
a flameproof enclosure
A common challenge for many elec-
tronic controllers located in hazard-
ous areas is setting the temperature
and alarm setpoints. Adjustments
normally require the power supply
to be switched off before opening
the flameproof enclosure housing,
which makes accurate, onsite set-
ting challenging, says the company.
By contrast, the EXTC 100 (photo)
is a fully adjustable, remote-con-
trolled, electronic controller that is
housed in a flameproof (Ex-d type)
enclosure. It can be fully controlled
by means of an intrinsically safe,
remote-control infrared beam —
without having to open or turn off
the unit. All settings are stored and
accessed securely, using a four-
digit password. This hazardous-area
Signal Fire Wireless Telemetry
thermostat is available for electric
trace heating circuits and other pro-
cesses. — Quintex GmbH, Lauda-
Königshofen, Germany
www.quintex.eu
Device supports wireless
monitoring of field assets
The Sentinal-RTD (photo) is the lat-
est interface option that can be
added to this company’s Wireless
Measurement Computing
Remote Sensing System, to sup-
port two-, three- or four-wire RTD
sensors. As part of the company’s
Class 1, Division 1 Sentinel node
family, the Sentinel-RTD provides a
direct connection to P100 2-, 3- or
4-wire RTD sensors (and other types
are available upon request). The de-
vice has a long life with an intrinsi-
cally safe battery pack, making it
especially well-suited to function reli-
ably in demanding outdoor industrial
environments (such as oil-and-gas
operations). The Signal Fire Remote
Sensing System (SFRSS) uses long-
range, mesh-networking technology,
and uses Sentinal (and other types
of) battery-powered wireless notes to
interface with the sensors. The nodes
extract and then communicate data
to control devices via a gateway. The
system has an open architecture,
which allows users to integrate many
types of sensors, including RTDs, to
Saelig monitor remote assets. In this way,
28 CHEMICAL ENGINEERING WWW.CHEMENGONLINE.COM
the SFRSS can be used to build a
full wireless network using the best
sensors for each sensing application
point. — Signal Fire Wireless Telem-
etry, Hudson, Mass.
www.signal-fire.com
Device manages high-precision
thermocouple measurements
The TC-32 high-precision, thermo-
couple-measurement device (photo)
offers 32 thermocouple channels
plus 8 digital inputs and 32 digi-
tal outputs/alarms. For larger sys-
tems, users can add the TC-32-ESP
module to double the number of
TC inputs (to 64), digital inputs (to
16), and digital outputs/alarms (to
64). The TC-32 and TE-32-EXP are
19-in.-rack-mountable and feature
easy-to-connect, mini-jack connec-
tors for all thermocouple channels.
The device features USB and Eth-
ernet interfaces, 24-bit resolution for
high-accuracy measurements, and
includes a wide range of software
support. It can provide digital alarm
notifications when a temperature
exceeds programmed limits. —
Measurement Computing Corp.,
Norton, Mass.
www.mccdaq.com
Multi-function loop calibrator
uncovers hidden problems
The portable PIE 850 Multifunction
Calibrator (photo) is a handheld instru-
ment that combines eight single-func-
tion calibrators (including a milliamp
current calibrator, a voltage calibra-
tor, a thermocouple calibrator, and
more), legacy 10–50-mA capabilities,
and advanced troubleshooting tools.
The PIE 850 is capable of diagnosing
common loop-control problems that
are hard to find, says the company.
Loop-diagnostic technology and the
LoopScope feature show key loop
parameters — such as current, volt-
age and resistance — at a glance. It
is compatible with all common instru-
ments, smart transmitters and PLCs
with 14 types of thermocouples and 9
types of RTD tempearature sensors,
says the company. This lightweight
tool includes a backlit display for
use in dark environments, a protec-
tive rubber boot and built-in stand, a
hands-free carrying case and all test
leads.— Saelig Co., Fairport, N.Y.
www.saelig.com n
Suzanne Shelley
JANUARY 2016
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/19 1/5
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/19 2/5
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/19 3/5
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/19 4/5
12/28/2018 January 2019

https://view.imirus.com/8rLRGj2mK8G/document/13064/page/19 5/5
 ANSWERS
INSTRUMENTATION, TUNING
Allan Kern, APC Performance LLC
Top 10 tuning tips
for control engineers
Best practices to follow to increase loop tuning performance and
reliability in a facility include increased efficiency, controller gain,
and being aware of loop interactions.
C
ontrol engineers often face ques-
tions and concerns about loop tuning.
Often, there are no readily available
answers, unless time is available for
a step test or to research a loop’s past
performance record.
By applying the following guidelines, a control
engineer can tell whether a loop is tuned appro-
priately or if it may be a source of performance
problems, and then make or recommend the
appropriate changes.
Note: Tuning is given in the format “gain/integral/
derivative,” where gain (or proportional control) is on
a percentage-of-span basis, integral (or reset) is in min-
utes per repeat, and derivative is in minutes. The term
“valve” may refer to any final control device.
‘
Level is an integrator and is often
nonlinear, so that proportional control
provides the most reliable overall
’
long-term performance.
1.
Don’t waste time on flow loops. Set the tun-
ing to 0.25/0.25/0.0 and move on. Spending
more time is usually pointless because process
gain will normally vary with valve position itself.
The relatively small gain—overall ideal flow control-
ler gain is 1.0—reflects operation at low valve open-
ing, where actual flow response is usually greatest.
At higher valve openings, the short integral time will
make up the difference, typically bringing flow to
setpoint in a matter of minutes if not seconds, which
22 • November 2018 CONTROL ENGINEERING
is usually much faster than the rest of the process.
This is often the most salient consideration when
tuning any loop—being timely in the context of the
process, while minimizing the potential for con-
trol instability. Large, slow valves may need a lon-
ger reset; occasional loops may need an even smaller
gain, especially if they’re operating at very low valve
opening.
2.
Don’t waste time on level loops. Set the
tuning to 1.0/response time/0.0 – and move
on. Response time, in level control context,
is the approximate time it would take for the ves-
sel, beginning at setpoint (typically 50%), to com-
pletely fill (or empty), in response to an average
disturbance, in the absence of any control response.
Typical refining process response times are 1 to 3
minutes for boots, 5 to 15 minutes for drums, and
30 to 60 minutes (or longer) for tanks.
Where severe disturbances are possible, or
where potential consequences of overfill or under-
fill are severe, use a larger gain instead of a shorter
integral (excessive integral action on level loops is
the most common source of process oscillations).
The level loop gain should rarely be more than 2.0.
This level tuning guideline applies whether or not
the level is cascaded to flow.
Level is an integrator and is often nonlinear, so
that proportional control provides the most reliable
overall long-term performance. There are several
common alternative strategies, such as gap gain, surge
control tuning, and dynamic optimization of vessel
inventory under model-based control. However, none
of these approaches add value or are more reliable.
3.
Don’t waste time on temperature, pres-
sure, composition, etc., unless they are
already cascaded to flow (or sometimes
pressure). If not cascaded, implementing a cascade
is the next best step to take. The cascade will lin-
www.controleng.com
 ANSWERS
INSTRUMENTATION, TUNING
earize the process response, thereby allowing opti-
mal reliable tuning across the full operating range.
Where a flow meter is not available, it may be pos-
sible to characterize the controller output based on
the valve characteristic; however, this technique
carries several uncertainties and vulnerabilities
and is a distant second-best practice.
4.
Invest time on pressure, tempera-
ture and composition loops that are
cascaded to flow (or sometimes pres-
sure). The cascade structure, in addition to lin-
earizing the process response, allows loop
gain to be accurately determined from pro-
cess data at two or more operating points.
Exercise: Try this calculation on a common heater
temperature/fuel gas flow cascade. (See link online.)
5.
Set gain to roughly ½ to ¾ of the
observed gain to maintain long-term
stable and reliable performance due to
Loop tuning cycle of rework
Sites frequently raise the priority of loop tun-
ing efforts, and allocate significant financial
and human resources, to remedy conditions
such as loops in manual, frequent citing of
loop tuning as a contributor to process upsets
or trips, to take advantage of the latest most
promising tools and technologies, to “re-engi-
neer” degraded multivariable controllers, etc.
Unfortunately, important considerations from a
process operation perspective are often over-
looked. Courtesy: APC Performance LLC
24 • November 2018 CONTROL ENGINEERING
potentially changing process conditions, which
can be expected to occur in most loops for many
reasons. Set the integral equal to the process
response time that is determined by using a step
test, process experience, or historical data—erring
on the long side. There is a significant long-term
performance benefit to a slightly smaller gain and
longer integral than traditional error-minimiza-
tion or quarter-amplitude decay criteria—such as
Zeigler-Nichols—will recommend, up to the point
of exceeding the process timeliness consideration.
6.
Do not use derivative action. Deriva-
tive is a way to reduce total error by using
more aggressive gain and reset, and then
relying on derivative to put on the brakes. This
is like speeding up and then braking hard as you
approach a stop sign, to get there slightly sooner.
Derivative action is inappropriate in most indus-
trial process control applications when observing
process speed limits, minimizing overshoot and
oscillation, and safeguarding process stability at all
times are paramount. Many modern tuning soft-
ware packages that routinely recommend non-zero
derivative settings are inexperienced. Model-based
multivariable control algorithms do the same thing
to achieve “error minimization” or “profit maximi-
zation,” but the resulting aggressive behavior, as in
single-loop control, usually needs to be backed out
to provide long-term reliable performance. This
caveat is in addition to the traditional concern
about large derivative bumps due to transmitter
noise or instability.
7.
Controller gain is heavily dependent on
span. For example, a pressure controller
with a span of 0 to 1,000 per square inch
gauge (PSIG) (using a modern smart transmit-
ter) will need a gain that is 10 times larger than
the same controller spanned for 900 to 1,000 PSIG
(as such loops were often designed in the past, to
have greater accuracy within the actual operating
region) to provide the same control response to
a given error. Without understanding this differ-
ence, there is reluctance to accept the larger, but
functionally equivalent, gain values. As a loose
guideline, gain is typically 1.0 for every 100 to 200
units of span for large-span temperature and pres-
sure controllers.
8.
Be aware of loop interactions. When the
action of one loop strongly impacts other
loop(s), the user will need to decide which
loop should be tuned normally and which one(s)
reducing gain and increasing reset time. There is
often no practical reliable option to tune all loops
in a set of interacting loops for timely response,
www.controleng.com
 ANSWERS
INSTRUMENTATION, TUNING
due to the technical demands of doing so, using
techniques such as “decoupling,” and the reality
of changing process gains, which undermines any
such schemes. The speed guideline for interacting
loops has a similar basis to the traditional cascade
rule, except in reverse.
For example, the secondary loop(s) should be
detuned to be at least three to five times slower
than the primary (high priority) loop. This funda-
‘
Model-based control remains subject
to most of the same tuning caveats as
traditional loop tuning, such as the effects
of variable process gains, the implications
of interactions, discretion regarding
feed forward, and leaving level controls
in the base-layer with robust tuning to
’
maximize process reliability.
mental rule has been historically overlooked and
ignored in multivariable control. This frequent-
ly results in unstable performance and ultimately
contributes to a degraded performance unless all
the involved models remain highly accurate, which
is rarely realistic.
9.
Be bold with gain and cautious with reset
when tight control is necessary. There is a
perception that too much gain may cause
cycling, and that since reset is in units of time,
a shorter setting might bring faster control. In
reality, proportional control action is instanta-
neous and too much reset, especially in combi-
nation with too little gain, is the most common
cause of process oscillations. So where tight-
er control is necessary, use more gain (up to the
limit of actual average process gain) and tune
integral more accurately (being careful not to
make it less than actual process response time).
10.
Understand when to use, and when
not to use, feedforward. Feedforward
can be advantageous when a major
26 • November 2018 CONTROL ENGINEERING
disturbance is well understood, when its model
(gain, response time, and deadtime) does not
change significantly in time for any reason, and
where it is warranted to avoid hitting a hard pro-
cess limit or to capture large earnings or avoid
large losses. When these criteria are not met,
especially if model dynamics (response time and
deadtime) are not reliably known, feedforward
should be avoided. Every feedforward model
comes with engineering, reliability, and mainte-
nance costs, which the wholesale use of feedfor-
ward in model-based multivariable control has
emphasized.
Successful loop tuning—defined as mini-
mizing rework, detuning, retuning, remodeling,
etc.—relies as much on an understanding of pro-
cess operation performance criteria as knowing
traditional single-loop tuning tools and methods
such as Ziegler-Nichols. When the process oper-
ation perspective is neglected, loop tuning often
is caught in a cycle of rework (See Figure), rather
than behaving as it should.
Another important takeaway is that mod-
el-based control remains subject to most of the
same tuning caveats as traditional loop tuning,
such as the effects of variable process gains, the
implications of interactions, discretion regard-
ing feedforward, and leaving level controls in the
base-layer with robust tuning to maximize pro-
cess reliability.
Originally, it was thought that model-based
control transcended most of these concerns, but
experience such as degraded performance, model
maintenance, and short lifecycle, has revealed
how these principles still apply.
Using these guidelines can help break the
rework cycle, increase success rate, and lead to
years of reliable maintenance-free process control
performance for most loops. ce
Allan Kern is an APC consultant with APC Per-
formance LLC. Edited by Emily Guenther, associate
content manager, Control Engineering, CFE Media,
eguenther@cfemedia.com.
M More
ANSWERS
KEYWORDS: Loop tuning, process optimization
Guidelines for loop tuning
Tips for loop tuning to avoid process issues.
CONSIDER THIS
What issues have you come across with loop tuning?
ONLINE
Read more about tuning online at
www.controleng.com.
www.controleng.com
 Special Report Process/Plant Optimization
W. SCRIVEN and A. MARTIN, Valero Energy Corp.,
Pembroke refinery, Wales; and D. SEIVER, Valero
Energy Corp., Houston, Texas

Utilize an optimizer to blend gasoline directly to ships

A new control system and single-blend optimization system This method required the refinery to maintain considerable
make it possible to blend gasoline directly into tanker ships for tank capacity. Tanks are costly, they require maintenance, and
delivery, rather than only into tanks, thereby saving millions in safety and environmental regulations require instrumentation
capital and reducing product giveaway. to prevent leaking, overfilling, etc. Additionally, the tanks are
The 550-acre Pembroke refinery (FIG. 1) has been operating never completely emptied, which represents essentially frozen
since 1964. In addition to its terminal equipment, the facility has assets—inventory that is never shipped but stays on the books
eight berths for ships and more than 100 tanks of various sizes and is taxed every year. The average tank inventory is generally
with total capacity of 10.5 MMbbl. The complex can facilitate 10 considered to be captured cashflow that is not realizable until
crude ships per month. the tank is emptied or the assets sold.
Touchups, even if they result in shipped product that meets
Gasoline blending system. The refinery’s gasoline blending specifications exactly, are also expensive. They result in poor
system has a capacity of 2,000 m3/hr. The blending situation utilization of both equipment and human assets; they put an
for the refinery is very complex, if for no other reason than the
number of different components and grades of finished gaso-
line it produces. The produced gasoline must meet the differ-
ent specifications for multiple markets (US, Europe, etc.). Most
outgoing products go by tanker ship, with a smaller amount go-
ing by road tanker.
Refining is a highly competitive commodity industry that
yields fungible products with low margins, and a refinery’s blend-
ing operation can be considered its “cash register.” This is the last
chance to optimize the final saleable product, as well as the fi-
nal opportunity to achieve as close to product specifications as
possible. Any giveaway at blending—caused by shipping better-
than-grade specification product—is lost revenue and cancels
out the benefits from optimizing upstream unit process areas.
As illustrated in FIG. 2, a small reduction in giveaway yields
impressive results; for a typical refinery, blend optimization FIG. 1. The Pembroke refinery in Wales has eight berths for ships and
could deliver millions of dollars per year in bottom-line savings. more than 100 tanks of various sizes with total capacity of 10.5 MMbbl.
It is estimated that the industry-wide total costs of Reid vapor
pressure (RVP) and octane giveaway exceed $2 B/yr in the US Crude Unfinished/components Product blends Finished product
and several times that total worldwide. Capturing even a fraction
of that opportunity can offer a strategic refining advantage. 1.2
Refinery margin, price factor × crude

Optimum
blending $$$$
Traditional gasoline blending methods. The original blend-
ing system in the Pembroke refinery combined set volumes of 1.1
multiple gasoline component streams in a header to produce Crude mixing
penalty Suboptimum
each batch of finished gasoline, which was then routed into a set blending
of tanks. A sample was taken to the laboratory for analysis. Af- 1.0
ter the results were determined, the mixture would be “touched
up” to meet requirements, if necessary, and a certificate of qual-
Crude receiving Process units Blending Product shipping
ity/analysis would be issued. Often, the touchup of an off-spec offsites
blend caused giveaway by adding a buffer to avoid carrying out
a second touchup. When the ship arrived, the blended gasoline FIG. 2. Capturing even a fraction of blend optimization opportunity
can present a strategic refining advantage.
would be pumped aboard from multiple tanks.
Hydrocarbon Processing | JUNE 2016 45
 Process/Plant Optimization
extra load/cost on the lab due to the retesting required; and,
if there is a problem and the blend testing comes out off-spec,
then shipments are delayed and incur potential demurrage
charges for holding up the ship.
In 2011, a study was undertaken to find ways to improve the
blending process: a project to replace the existing blending ar-
rangement with one that would allow the blending of gasoline
directly to ship rather than solely to tankage. Such a system had
the potential to save millions of dollars each year by reducing
the number of tanks needed in gasoline service, as well as satisfy
the key performance indicators of lower average blend giveaway
and the minimization of touchup blends.
Upgrading the blending system. The refinery was already
scheduled for a control system update, which would ultimately
include replacing the existing, and obsolete, distributed con-
trol system (DCS). The gasoline blending optimization proj-
ect was approved, and the project teams realized that it would
integrate well with a new DCS system upgrade in the blending
and shipping areas.
Each batch of finished gasoline is made by simultaneously
blending up to 21 different component streams in a header—
typically 10 or fewer streams at a time, with mid-blend pump
and tank changes. Each component has a flowmeter, a propor-
tional integral derivative (PID) controller and a target ratio set
by the control system. A set of online analyzers provide feed-
back for optimization and certification using state-of-the-art
redundant sample compositors. The system can achieve up to
20 quality limits simultaneously, although the number is usu-
ally closer to five. The output of the blending system (finished
gasoline) can go into tanks or directly onto a ship.
The blend optimization system used at Pembroke is com-
prised of a highly standardized mix of both hardware and
software components. The system is built of several layers: At
the bottom (just above the field devices) is the DCS; on top
of the DCS is an advanced regulatory software package, which
handles the ratio control functions. Above this layer is a blend
optimization supervisory system (BOSS) software package. Sit-
ting at the top of the pyramid, it processes data from analyzers,
Refinery blending unit DCS Blend optimizer
SBO blend SBO blend
FT
Tank AE M recipe ratio property
component A
M control control
Multi-blend
FT planning
Tank AE
M
M software
component B
FT
M VGBOP
Tank AE
M Redundant FTIR and
component C vapor pressure online
field analyzers
FT
M
Tank AE
M Multi-blend PT
component D planning
Coriolis software
meter
FIG. 3. A schematic drawing of the various parts that comprise a blend optimization system.
46 JUNE 2016 | HydrocarbonProcessing.com
flows, etc., and calculates a recipe that is optimal for the blend at
any given moment to meet the optimization objectives. The op-
timized recipe from the BOSS is then fed to the regulatory soft-
ware package, which calculates the appropriate flow setpoints
for the individual blend component control loops. For example,
it takes a 41% alkylate recipe and translates that into a setpoint
for the alkylate flow control loop. It performs other regulatory
level functions as well, such as starting and stopping pumps, etc.
FIG. 3 shows schematically how the different parts fit together.
Consistent data helps obtain operator acceptance.
A critical success factor of the new system is the presentation of
consistent views to the operator. The design and implementa-
tion of such a system must revolve around a single-screen inte-
grated approach to minimize the different operator system inter-
actions. All the blend regulatory controls—start/stop, routes,
ramp, ratio and pacing—are handled in the DCS via commands
from the blend optimizer, akin to a master/slave control rela-
tionship. The operator displays are kept to a minimum and are
integrated with the DCS graphic, which helps considerably in
obtaining operator acceptance of the new system.
The optimizer screen is the interface that the operators use to
start and monitor the blends. This screen is typically provided on
a separate interface screen, but using RDP from the DCS made
it possible to display this screen within the DCS system, simpli-
fying it for the operators and demonstrating the “single-screen
integrated” approach the project team and operators wanted.
Challenges encountered. As expected, designing and imple-
menting the new system was not without its challenges, many
of which coincided with systems not always being available
when needed. At the front-end engineering and design (FEED)
phase, there were gaps in the system knowledge, which is often
the case in projects where very old systems are in place and little
or no good documentation still exists.
To in-line blend directly to a ship, a composite sampler that
can take samples at required rates for the size of the blend and
provide a composite of exactly what went onto the ship (it is
very difficult to test the separate tanks on a ship) is required. The
sample compositor takes a cross-section of
the final product exiting the blend header.
This can be tested in the lab and a certifi-
M
cate of quality/analysis can be produced.
These custom-built, redundant-sam-
ple compositors, which are a proxy for the
M blend tender going to the final destination
(i.e. ship), must be very robust and very
accurate. The reliability of such a system
M is critical, as any failure would result in
Pipeline uncertain specifications of the gasoline
product loaded onto the ship. Note:
M
Backup procedures are in place that allow
for additional testing and sample collec-
tion, should the compositors fail to avoid
M Finished any off-spec loading to the ship. Establish-
product tank
ing the reliability of this system was a mi-
nor challenge, but in the end it proved to
be extraordinarily valuable to the overall
success of the project.

Another challenge came from the limited time available for
the factory acceptance test (FAT). There were delays in getting
all parts of the project approved due to the compressed tim-
escales needed to take advantage of the earlier-than-planned
DCS upgrade. Consequently, this reduced the time available
for the FAT. To make things worse, the system was being mi-
grated from one that had been in place for decades. Many pages
of leftover custom code needed to be identified and stripped
out before the new system could be fully programmed.
Hardware problems also had to be overcome. Communi-
cations with some of the antiquated and poorly documented
serial devices in the facility proved difficult. Some of the flow-
meters on the blend header had little or no parts availability
and documentation was sparse, making installation and tun-
ing impossible. It was decided to replace a number of these
instruments on the blender to improve long-term reliability.
Custom modifications to the DCS input modules were in-
stalled to integrate these into the DCS so that future updates,
tuning and maintenance would be simplified and could be car-
ried out in-house.
Results. The new system was first used to produce optimized
blends to tankage, and within weeks optimized blends were be-
ing successfully loaded directly to ship in parallel to loading from
existing tanks. Several months later, the site completed its first
100% full blend to ship, a parcel of 46 Mm3 taking over 50 hours
for an export cargo. All direct-to-ship blends tested to spec on
shipping, and again when received at their final destination.
As discussed, one of the eight gasoline storage tanks (used
previously for final product containment prior to ship loading)
was removed from gasoline blending service and reallocated to
another component system. This move freed up several mil-
lion dollars of gasoline inventory, mitigating required upgrades
and allowing tank scheduling decisions to be made on decom-
missioning or changing of service. Far fewer reblends are seen,
and the blends are closer to spec than before, which reduces
giveaway costs.
To maximize the system and manage assets effectively, a
multi-discipline team was developed to monitor, maintain
and improve the gasoline blender and its peripheral systems.
A series of key performance indicators (KPIs) was set up, and
the site began monthly tracking of the optimizer savings and
performance. To date, the results have shown excellent per-
formance, and the new system provides more flexibility for
product planning and scheduling. The site has been able to
maximize the value of the gasoline blender by reducing the
time it takes to fill, test, certify and load multiple tanks to ship,
resulting in increased system throughput and, at the same time,
minimizing the cost of finished product by blending the least
costly components required to make on-spec gasoline.
System troubleshooting has also become much easier. For
example, the old system had four graphics for the operators to
monitor pumps and controllers, whereas the new system con-
densed that down to one graphic. Operators can now quickly
view the blend header graphics, which are more of a graphical
representation of the system, and provide an overview of all
running pumps, control valve positions and flowrates. A blend
analyzer screen (FIG. 4) provides an expedited way of checking
gasoline blend property value.
Process/Plant Optimization
FIG. 4. A blend analyzer screen shows an overview of all running
pumps, control valve positions and flowrates, providing an expedited
view of gasoline blend property values.
It is much easier to see the sequential function charts execut-
ing the sequencing of the startup and shutdown with the new
displays; if there is a problem, the operators can detect its ori-
gins and how to correct it. It is also useful when calibrating new
turbine meters, making it possible to see exactly what K-factors
have been installed and the changes needed for calibration.
Additional benefits may be considered, such as the reduc-
tion of approximately 70 Mbbl of gasoline, which reduces the
onsite storage of hazardous materials, the need for operators
and maintenance to work around a commissioned tank, and the
environmental risks of hazardous material storage. Overall, the
project team identified process improvement opportunities, de-
livered one-time capital benefits of several million dollars, and
provided a sustainable blending optimization solution that has
already contributed significant ongoing annual operational sav-
ings and is expected to continue for many years forward.
WILL SCRIVEN is a senior control system specialist for the
blending and shipping area at Valero Energy Corp.’s
Pembroke refinery. He has six years of experience in the
control systems department. Prior to this role, Mr. Scriven
completed an apprenticeship in instrumentation and was
an instrument technician at the site for two years.
ANDREW MARTIN is a senior process engineer with Valero
Energy Corp. He has five years of experience as an engineer
in the Pembroke refinery as part of the process engineering
and advanced control teams. He also spent three years in the
power generation sector, where he was a performance and
commercial engineer as part of the operation and
commissioning team for the Pembroke combined-cycle gas
turbine (CCGT) power station. Mr. Martin has developed advanced control
applications for steam and utility plants and led process optimization for
gasoline blender and CCGT power station operation. He has experience in power
station commissioning, energy markets and multiple DCS platforms. He holds
an MS degree in chemical engineering from Cambridge University, where he
attended Queens’ College.
DAVID S. SEIVER is the director of blending APC technology
for Valero Energy Corp., managing the rollout of the
company’s single-blend optimization (SBO) program at its
refineries. Mr. Seiver has over 25 years of experience in the
petrochemical and industrial gas industries. Prior to coming
to Valero, he specialized in gasoline and diesel SBO for
ConocoPhillips at its Wood River refinery, as well as provided
technical blending assistance to other sites. He has extensive expertise in APC,
SBO and NIR modeling, and has two US patents in the APC field and one US
patent for gasoline optimization. Mr. Seiver holds an MS degree
in chemical engineering from the University of Houston.
Hydrocarbon Processing | JUNE 2016 47