r

Engineering Practice

Spreadsheet Calculates
Critical Flow
Sunil Kumar
Worley Engineering
This method encompasses awide
range of isentropic and frictional
factors, and subcritical flow
hen handling gas, we usu- ate between choking and non-choking that have not been widely publicized.

W ally encounter two types of

flow. The most common one
is non-expanding flow. This
is the normal condition in most oper-
ating process lines. Generally, the
flows and to estimate their magni-
tude. This article provides an Excel
spreadsheet to put on the computer,
address expanding-flow conditions,
and determine flowrates without
Here are some:
• Two inputs have discrete and limited
ranges: K from 5 to 90 and 'Y from 1.0
to 1.4. This raises the need for un-
easy extrapolations and interpola-
flow is isenthalpic. The impact of a ve- hours of research. tions. We will show that 'Y and Kare
locity change is minimal and assumed key variables because YCR and
to be constant throughout the length Specific constraints (t:.P I PoJcR are strongly influenced
of the pipeline. Any pressure decrease We focus upon expanding flow through by them
is attributed to frictional losses in the constant diameter pipes. If you have • Reading a plot is an imprecise exer-
line. If the total pressure drop is less an orifice, see Reference [2). The first cise. The accuracy of the data is sub-
than about 10% of inlet pressure, the item of business is to ascertain the ject to the precision adopted by the
unmodified Darcy equation [J], can be thermodynamic condition of flow. It is user. Moreover, we find that the
applied to calculate the flowrate with either isothermal or adiabatic. user tends to pick heavily conserva-
reasonable accuracy. In general, for short pipes (less than tive values, and accordingly the re-
By comparison, expanding flow is the 10 m), the assumption of adiabatic sultant flow gets unnecessarily
engineer's headache. This worrisome conditions is reasonable. Gases have a overestimated.
occurrence starts from an abnormal poor heat-transfer coefficient. We can
flow condition, which manifests itself thus, furthermore, picture a massive Thermodynamics overview
by a very high pressure drop. Flowrate flow (approaching sonic velocity) in a The approach used in this article is
calculations have to account for the ef- relatively short pipe. purely thermodynamic and rational
fects of adiabatic expansion and friction For longer pipes, the assumption of [5] . No experimental data or empirical
losses at high velocities. Here is a sam- an isothermal condition is reasonable. correlations have been used. The goals
pling of abnormal situations: However, with pipes shorter than 100 of the equations and the resultant
• At the tailpipe of high-pressure m, the realistic flow condition leans spreadsheet are to:
blowdown systems during shut- more toward adiabatic. Therefore, for 1. Check the status of flow. It is either
down or depressurization; seen par- the purpose of deriving critical corre- critical or sub-critical.
ticularly on offshore petroleum and lations within this article, adiabatic 2. Determine (AP/P(YCR•
gas production platforms conditions have been assumed. 3. Determine the flowrate.
• Accidental gas-blowby as a result of The flowrate can be determined When a compressible fluid accelerates
inadvertent opening a drain valve of using appropriate flow equations, ei- through a restriction (such as a pipe,
a pressurized vessel, or stuck-open ther the Darcy Equation [1) or the orifice, venturi or nozzle) as a result of
failure oflevel-control valves Robert Kern Equation [3). If critical massive pressure reduction down-
• Rupture of heat exchanger tubes flow is the status, one way to deter- stream, the velocity and specific vol-
• Bursting of a rupture disc mine flowrate is from the Crane cor- ume increase at the expense of en-
• Opening of a pressure safety valve relation plot [4). The Crane method thalpy. The relevant thermodynamic
If you are going to plan mitigation pro- simply uses the Darcy formula modi- equations for the state of the gas follow
cedures for abnormal events, or to de- fied with an expansion factor YCR a reversible path, but with reduced
sign appropriate discharge systems, a [1]. This expansion factor takes into isentropic efficiency. Reduced isen-
pre-check of flowrates is needed for account all modifications in flow re- tropic efficiency is approximately the
the design record. It is not easy to fig- quired as a result of variations in sum of frictional losses and heat losses.
ure this out in a realistic frictional density and velocity. The Crane plot We start with a basic energy equa-
system, when the flow approaches the correlates YCR, t:.P IP and K. (see tion and develop from it an equation for
choked (critical) condition at an end- Nomenclature Table on p. 63 for overall critical-pressure ratio. Since
point. The equations are incredibly symbol definitions). However, the there are a plethora of subscripts, the
complex. However, we have comput- Crane method, being a plot-based entire flow system has been split into
ers and so it will be easy to differenti- correlation, poses a few limitations nodes. They are shown in Figure 1.
62 CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002
 19-24 May 2003 • Frankfurt am Main

WORLD FORUM Main exhibition sectors

Research and Innovation
Engineering, Plant Design and Construction
FOR THE Mechanical and Thermal Processes
Laboratory and Analytical Techniques
Instrumentation, Control and Automation
PROCESS Pumps, Compressors, Valves and Fittings
Pharmaceutical Technology
Packaging and Storage Techniques
INDUSTRI.ES .__ - Mater,iD Technology
Industrial and labour Safot1t
Biotechnology -------
Environmental Protection

Special Show
e-Services for the Process Industries
Meet more I an ,000 exhibi1
and over 200,000 visitors f .., International Congress
~ AUJ.m..Qre than 800 lectures
from all over the world and~ ~

Please send me further information on ACHEMA 2003

Organizer:
DECHEMA e.V. D Documents for Exhibitors D Visitors Information
P.O. Box 15 01 04
D-60061 Frankfurt om Main
Fox ++49 (0) 69 7564-201 Nome
Exhibitor Information Coll Company/Institution
Tel. ++49 (0) 69 7564-230/ 238/-434
Visitor Information Coll Street, P.O. Box
Tel. ++49 (0) 69 7564-249/ -242 Postcode/City/Coontry

www\ he Phone

E·l.\oil
fox
 NOMENCLATURE Critical flow
conditions
A = Internal cross-section area of u, = Velocity at inlet tls of the pipe P2,V2,U2
pipe, m2 (start of Row), m s P 1•V1, U1
D = Inner dia. of pipe, mm U2 = Velocity at outlet tip of the pipe,
d = Infinitesimal differential operator, m/s Maximum
dimensionless Us = Velocity of sound in gas medium length 100 m
ft = Moody friction factor, m/s Ambient
dimensionless Vo = Specific volume at upstream stag- pressure PA
F = Frictional head losses, m2/s2 nation point, m3/kg
K = Resistance coefficient, v, = Specific volume at inlet-tip of the
dimensionless pipe (start of flow), m3/kg FIGURE 1. Consider a small diameter
= Length of flow pipe, m pipe between an inexhaustible high-pres-
V2 = Specific volume at outlet tip of the sure source and a large ambient receiver.
m = Slope defined in Equation (19), pipe, m3/ kg
dimensionless It is possible that the flow from the pipe
W: Mass flowrate (critical), kg/h will approach the speed of sound
M1 = Mach Number at inlet tip of the
pipe (start of flow), dimensionless X
= Value of subcritical (t.P/PoJsuB
dimensionless '
M2 = Mach Number at outlet tip of the
pipe, dimensionless y = Equation 18), dimension ess
Value of Y correspondinr, to x (see
Po = Stagnation upstream pressure, (7)
kPaa YcR = Critical exkansion factor,
dimension ess
When dealing with adiabatic expan-
P1 = Pressure at inlet tip of the pipe sions, the fundamental PV relation-
(start of flow), kPaa t,. =
Finite change operator,
ship is Equation (8). Solve Equation
P2 = Pressure at outlet tip of the pipe, dimensionless
kPaa (t.P/Po)cR = Overall critical differential (8) for V, and write that into Equation
PA = Ambient pressure, kPaa pressure ratio, dimensionless (7) to get Equation (9).
r = (P 1/ Po)cR = Overall critical -y = Ratio of specific heat at constant
pressure ratio, dimensionless pressure to specific heat at pyr =P1V{ =constant (8)
uo = 0 =Stagnation upstream constant volume at upstream
velocity, m/ s conditions, dimensionless
2A: [[ p½ }P+2J du + f(tr dl) = o
• Node 0 is within a stagnant up- and transforms simultaneously into W P, P{YV1 "• u l, D . (9)
stream system, kinetic energy (t:..u2 I 2) as the flow
• Node 1 is the starting point of flow tends to stay isentropic. However, the Consistent with Figure 1, P 1 and V 1
• Node 2 is the discharge tip of pipe entropy in fact increases because of ·are constants. They are the upstream
into the ambient atmosphere, which velocity-induced friction. Equation (1) conditions. Taking the constants out
is at a low pressure is a fundamental energy balance [5) of integration, we get Equation (10).
In real situations, any expanding flow applicable over a fraction of pipe
emanates from an inexhaustible source length, 6.1: 2A
2
w2pfrv
1
?(p,Yr'rLP+2"•J u
du+ J(rr dl) = 0
P, 1, D
such as a vessel, or a large-bore pipe. As- Vt.P + t.u2 + t.F' = 0 (1) 1 1 (10)
sume that the u pstream pressure Po is 2
both constant and stagnant. At the Equ~tion (2) is an expression from Integration gives us Equation (11).
source conditions (Node 0), the assump- Reference [6). 2 (11)
tion of no-flow is reasonably valid. 2A ( y )
w2pYrv y+ 1 X
At the upstream tip of the pipe, the t.F = (rr ~) u; (2) 1 1
flow starts abruptly with some initial
velocity u 1 . The pressure Po has Plug Equation (2) into Equation (1). [p27 _p7]+2ln(~:)+fr(l~-li)
1
o
dropped to P1 as a result of an in-
crease in kinetic energy. As the flow Vt.P+ t,.u2 +(frt.l )u2 = 0 We can factor out P1 raised to the
2 D 2 (3)
progresses down the pipe, the thermo- (-y+1 1-y) power. from the first section
dynamic changes and frictional losses, In differential form, as the deltas of the equation. After doing some nec-
will further tend to drop the pressure shrink to infinitesimals, we arrive at essary algebra, Equation (11) trans-
until the critical pressure P2 is Equation (4). forms into Equation (12).
reached at the downstream tip of pipe (4)
(Node 2). The pressure P2 represents
the critical pressure at the tip, rather
VdP +udu+(fr~)u; = 0 ;P1A y
W Vi(Y+ 1) P1
2
[(P2)7 -l]+21n(u2)+frl~ l =)O
U1
2
than the actual downstr eam ambient Now, divide through Equation (4) by
pressure. We now have a criterion for u2 !2. The Mach Number and the resistance
evaluating critical flow. · coefficient (K) have been defined in Ref-
(5)
• If PA < Critical P2 , choking exists erence [5]. Use the definitions as in
• IfPA > Critical P2, there is no choking Equations (13) and (14). Substitute into
Substitute: u2 = (W2V2 IA 2) for flow Equation (12) and get Equation (15).
Energy balance in a constant cross-sectional pipe, to (13)
We next develop equations correlating get Equation (6).
P 1 and P2 . In an adiabatic flow
process, there is no heat transfer, and
2
2A dP
w2 v
+2du
u
+(t:r dl)
D
=O (6)
(14)
the shaft work is generally zero. As
Integrate Equation (6) between
the flow expands in the direction of de-
creasing pressure, enth alpy (the total inlet (Node 1) and the outlet (Node 2) 2 [ (PJ
r(y+l) P2 2'.:!:! ] +2ln u2 +K - 0
(uJ (15)
r -1
energy, the product V~) decreases conditions of pipe.
CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002 63
 TABLE 1. LIMITING CRITICAL VALUES
K 'Y =1.2 'Y =1.3 'Y =1.4 'Y =1.5 'Y =1.6
( AP/P0 )c11 YcR ( AP/Po)c11 YcR ( AP/Po)c11 YcR ( AP/P0 ) c11 YcR ( AP/Po)c11 YcR
1 0.62 0.52 0.64 0.51 0.66 0.50 0.68 0.50 0.70 0.49
2 0.64 0.54 0.67 0.53 0.69 0.53 0.71 0.52 0.73 0.51
3 0.68 0.58 0.70 0.56 0.73 0.55 0.75 0.54 0.78 0.53
4 0.7 1 0.60 0.74 0.58 0.76 0.57 0.78 0.55 0.80 0.54
5 0.74 0.61 0.76 0.59 0.78 0.58 0.81 0.56 0.82 0.54
6 0.75 0.62 0.78 0.60 0.80 0.58 0.82 0.56 0.84 0.55
7 0.77 0.62 0.79 0.60 0.81 0.58 0.84 0.56 0.85 0.54
8 0.78 0.63 0.80 0.61 0.83 0.59 0.85 0.56 0.86 0.54
9 0.79 0.63 0.81 0.61 0.84 0.59 0.85 0.56 0.88 0.54
10 0.80 0.63 0.82 0.61 0.85 0.59 0.86 0.56 0.88 0.54
20 0.86 0.64 0.88 0.61 0.89 0.58 0.91 0.55 0.93 0.52
30 0.88 0.64 0.90 0.60 0.92 0.56 0.93 0.53 0.94 0.50
40 0.90 0.63 0.92 0.59 0.94 0.55 0.94 0.52 0.96 0.48
50 0.92 0.63 0.93 0.59 0.94 0.55 0.95 0.52 0.96 0.47
60 0.92 0.62 0.94 0.58 0.95 0.54 0.95 0.51 0.97 0.46
70 0.92 0.62 0.94 0.58 0.95 0.54 0.96 0.50 0.97 0.46 '
80 0.93 0.62 0.94 0.57 0.96 0.53 0.96 0.49 0.97 0.45
90 0.93 0.62 0.94 0.57 0.96 0.53 0.96 0.49 0.97 0.45
100 0.94 0.61 0.95 0.57 0.96 0.52 I 0.97 0.48 0.98 0.44
Columns 2, 4, 6, 8 and 10 show the variation between (APIP0 ) CR, 'f and K.
Columns 3, 5, 7, 9 and 11 show variation between YcR, 'I and K.

We can remove pressures and veloc- the pipe's endpoint. This is equiva-
TABLE 2. PROBLEM SETUP
ities by using thermodynamic ratios. lent to the line's flow velocity. Now
Input item Value
The math is rather involved, so it is estimate flowrate based on pipe di-
condensed in a box on p. 65. Substi- Upstream pressure, P0 6,600 kPaa
ameter. This approach assumes adi-
tute the critical ratios from Equations Downstream discharge
abatic expansion. Use Equation (45) pressure, PA 200 kPaa
(43) and (44) into Equation (15) to ar- • Or, preferably, use the isothermal Upstream specific
rive at Equation (16). correlation [5], which is conserva- volume, V0 1/58 kg/ml

2 [
{y+l}Mf 2 l
Mf(y+l) (2+{y - l}Mr) - +
~· ] (16)
tive, and usually gives a flowrate
20-30% higher than the adiabatic
calculation
• If the flow happens to be subcritical,
lsentropic coefficient, "I 1.55
linner dia. of pipe, D
Length of pipe
Number of e lbows
49.2 mm
100 m
4
21n /2+(y- l)Mf +K =O (as determined by the preceding two Resistance coefficient K
(y+l)Mr steps) use the modified Darcy for- (as estimated for above 45
mula with appropriate YCR. Skip to 49-mm pipe data)
Useful critical ratio the last part of this article for fur-
Equation (16) represents the flow sta- ther discussion ginally higher. The reason for the de-
tus during endpoint choking between Equations (16) and (17) are computer- viation in results is attributed to the
Node 1 and Node 2, and as such, it is ized in the spreadsheet that was limitation of maximum 'Y = 1.4 in the
able to provide a solution for part of promised at the beginning of this article. Crane method, whereas, for the ac-
the critical flow problem. This equa• It is provided as a download at tual gas 'Y = 1.55.
tion is a complex function of two vari- www.che.com/CEEXTRA. In addition to the spreadsheet pro-
ables, M 1 and K. For applications, en- The derivation of Y CR is shown in vided, we pi;esent Table 1 for a large
gineers are happier correlating the box on p. 66. The estimation of range of conditions so that readers can
between Node O and Node 2. We need YCR can be accomplished with the fol- verify the accuracy of the program after
more manipulation to have a solution lowing calculation steps. downloading. Table 1 indicates a wide
that estimates (Pzl Po)CR· • Estimate K for the pipes, and 'Y at variation in the critical values with re-
There is a way to estimate the pres- Node 0 spect to 'Y and K. Crane publishes a sim-
sure ratio between Node 1 and Node 2. • Determiner using the Excel spread- ilar table in Crane A-22 [4]. Figure 2
Multiply Equation (32) by Equation sheet method, as described shows the variation between
(44) and thus come up with a correla- • Substitute r into Equation (52) to (t.P IPo)cR, 'Y and K. Figure 3 shows
tion for (Pzl Po)cR: Equation (17). determine the value of YCR variation between YCR, 'Y and K. Both of
the charts indicate a reasonable varia-
r-(p
- R
2) _
-
[o.5(y +l)Mr]½
r(r+1)
Putting the results to work
We present an example for doing
tion with respect to isentropic factor.
Here is the example. Estimate the ma-
o CR [1 +0.5(y- l)Mrj2(r-1) (17) choked-flow calculations. Equation (17) terial loss for a tube-mpture incident in
.is used to determine the maximum a double-tube gas heater. The gas heater
The simultaneous solution of Equa- critical (choked) flow as well as to esti- is receiving gas from a 102-mm main
tions (16) and (17) eliminates M and mate the critical expansion factor. gas-line via a 49-mm branch line. The
yields a value for r, the critical pres- We solved the example by two length of branch line is approximately
sure ratio. We can now determine criti- methods, the Crane plot-correlation, 100 m. The upstream conditions of gas
cal flow in either of the following ways: and the spreadsheet. The flowrates and other relevant data are in Table 2.
• Determine the velocity of sound at from Crane plot-correlation are mar- The estimate ofK =45 comes out of the
64 CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002
 Engineering Practice

- y 1.2
- y 1.3
- y 1.4
- y 1.5
- y 1.6

FIGURE 2. The critical pressure difference is influenced by 'I, FIGURE 3. The critical expansion factor is influenced by 'I,
especially at modest resistance coefficients especially at low resistance coefficients

FRICTIONLESS THERMODYNAMIC RATIOS

t is a bit complicated to move from Mach Between Node O and Node 1 with Mach Numbers as derived from Equa-

I Number to the general isentropic flow Apply Equation (26) to the pipe inlet (Figure tion (21 ).
correlations. But it is necessary, so here 1). The subscripts are O and 1.
are the steps. 2 2
2
(27) M , = l ,,c,,v,
u¾ (35)
The Mach Number in a flow is defined as _r_ (PoVo) +uo =_r_ (P1 V1) +!2. ,,
a ratio between actual-velocity of gas to the
r- 1 2 r- 1 2
There are two velocities: uo = 0 and (u 1)2
(36)
velocity of sound, at the fluid conditions
prevailing at that particular point. [5]. = 'IM12P1V1 (from Equation ((20]). Plug
The velocity of sound in a gas medium is: these into Equation (26).
(37)
u, =ffe (20)
-Y- (PoV,o) = - Y (PV ')Mf-P1-
1 1) + - V1 (28) Rearrange Equation (37). Equation (38)
Thus, the Mach Number is: r -1 r-1 2 shows the velocity ratio.
Rearrange and simplify.
M=.j; V (21)
u,_
[f2+(r - l)M?}Mrl
_ P1V1 2
There is no motion at a point. Eliminate (29) Uz - 1 {2+(r - 1)Mr}M~
friction from Equation (1) and rewrite as PoVo - 2+ (r - 1
)Mr (38)
Equation (22). In thermodynamic text· From Equation (8), the fundamental adia- Other ratios, such as V1/V2 and P2/P1
books, this is the fundamental energy bal- batic equation, P1Vl'Y = PoV o-Y [5] can be can be worked out with some basic thermo-
ance for frictionless isentropic flow. written in the form o Equation (30). dynamic knowledge [5]:
2 (22) In pipe flow:
Vt,.P+ 0.5t!.u = 0 V1 = (Po)f
Integrate from general Point 1 to general V0 P1 (30) {u/ V) = constant (39)
Point 2: Substituting the above into Equation (29):
v, _ [{2+(r - l)Mn Mrl (40)
2

I
2
JVdp + 0.5Judu = O
I
(23)
(
P1
Po
)7-_2+(r- 1)Mr
2 (31) ½ - ~ {2+(r - 1)Mr}M~
The isentropic process has a well-known In adiabatic flow:
rela tionship [7]. Solve for P,/Po: PVY= constant (41)

fVdp = _L
l y- 1
(P2Vz - P,V1) (24) P
p~ 2
II
= 2+(r - 1)Mr
l*' (32)
f)_ _ [{2+(r - l)M~}Mrl
P, - {2+(r - 1)Mr}M~
r
2
(42)
Substitute Equa tion (24) into Equation Between Node 1 and Node 2
(23); integrate the seco nd term, and write Apply Equation (26) at the pipe end. The If the flow is choked, then there is sonic
Equatio n (25). subscripts are 1 and 2 flow at the pipe outlet. In short, M2= 1. This
25 r u2 r ) u2 creates special ca ses of Equation (38), (40 )
r~/P2½- P1V1) +(uJ _uJ) =o 1 1 r - / P1V1)+-f =r -l (P2½ +; (33) a nd (42) a s follows:

Because of the zero, it is possible to convert

Equation (25) into a general, fundamental,
Rearrange Equation (33):
(~;)CR=(~;)CR =I2t(; ~~~f (43)
frictionless, isentropic-Aow correlation
2
(26)
u2[
1 (r2)'-1'JV,
1)uf
+ 1]-
u2[ 2-yPzVz + ,]
- 2 (r - 1)u~
(34) r (44)
_r_ (P V ) + !£.. = constant
r -1 2
Replace the velocities inside the brackets (~tR ~[2t(;~~~fr 0

CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002 65

 ~ - - -- -- -- -- - -
OVERALL CRITICAL EXPANSION FACTOR, YCR USING THE SPREADSHEET
n general, the Excel spreadsheet solves

D
nee the critical pressure ratio is known,
the mass flowrate at choking can be
estimated.
rearranging the terms yields Equation (47).
W = 0.1265D2 x
,--
[ ____ ] _ _ _ (47)
I complex equations where a trial-and-
error method is required. Equations
(16) a nd (17) are non-linear and simul-
Adiabatic approach: Po x(l [P2J2J taneous equations with two unknown
variables: M1 and r. Therefore, in order
The gas velocity at the endpoint of the pipe ' 2Vo{K +
is the same as the velocity of sound at end
21n[P2~0J} -
/ 11
Po to determine the value of r, both equa·
lions must be solved simultaneously, to
point conditions. If we know the inner di- Substitute the pressure ratio consistent eliminate M 1•

l
ameter of the pipe, it is possible to trans- with Equation (17); defined as P2/Po = r
form the end point conditions to upstream Here are the steps inside the Excel
spreadsheet method. We have appended
conditions. This is the adiabatic approach, 2
and we use appropriate units and the criti- W = 0 1265D2 P.o[ I- r ] the program in www.che.com/CEEXTRA
cal pressure ratio to develop this correla- with three notational changes. While
·
1 [
N o{K + 21n[frl} (48) working in Excel, the M 1 has been re-
tion , Equation (45).
placed with M, -y with g, and P2/Po with r.
y+ I Equation (49) is Darcy formula [ I], 1. Define Names K, g (for -y) and M (for
0.089D2 r(Yvo)rr (45)
0.1265D 2YcR✓(pw:2 )
W = M 1) in separate cells that refer to their
W= (49) respective values. Note: Use Name/De-
Isothermal approach: fine command on Insert menu.
However, the use of isothermal Row equa- 2. Print the following formula in a cell
Substitute P2/Po = rand rearrange Equa- which refers to the Left hand side of
tion is more conservative, and yields about tion (49).
30% more Row than does the adiabatic Equation (16) and returns the value to
method. In addition, as discussed in the main zero via "Goal seek" command as ex-
W = 0.1265D2YCRt o~r) plained below.
body of the article, the Row stays somewhere (50)
between isothermal and isentropic. There- =2/(M"2*(g+ 1))*(((g+ 1)*
Equation (48) equals Equation (50).
fore it is reasonable to use the isothermal M"2/(2+(g- 1)*M"2))"((g+ 1)/2)-1)
Solve for YCR· +2*LN(SQRT((2+(g- 1)*
equation for the estimation of mass Rowrate.
The mass Rowrate for expanding Row is K( 1- r 2 ) M"2)/((g+ 1)*M"2)))+K
given by the following formula [5], which we YcR = (51) 3. Print the following formula in a cell that
call Equation (46). The usage of this expan- 1{ K+ 21n[fr])(1- r) refers to the right hand side of Equation
sion factor is limited for Rows only through
We can cancel the term ( 1-r) from the nu- (17) a nd returns the value of P2/Po.
constant-cross-section (diameter) pipes.
merator and denominator. We get Equa- =((g+ 1)*M"2/ 2)"(g/ 2)/(1 +(g- 1)*
tion (52) for critical expansion factor. M"2/2)"(g*(g+ 1)/ (2*(g-1)))
2
W- [ A Po ] ( ~) (46) K(l+r) 4. Use "Goal seek" command (from
- ~ K +2/n[¾ ] Po
YCR = 1 { K+ 21,{fr])
(52) Tools menu) to calculate M, by having
the M-value to change so as to set the
Converting this equation to convenient value of the left hand side of Equation
(metric) units, substituting for area and den· NOTE: For flows through an orifice, the ex- (1 6) to zero.
sity (a s defined in the nomenclature), and pansion factor is provided in (2). • 5. Print the solved values for rand M.
6. Further, the value of YCR can be deter-
mined by a simple formu la inserted in
Crane book [6]. The Moody Friction Fac- "I and K. We achieve a less-conserva- spreadsheet (or done on a desk calcu-
tor is 0.02 for 100,000 mm of pipe to- tive and more-precise flowrate. Also, lator). The spreadsheet formula for
gether with 4 standard elbows and exit the Excel model is a lot quicker . YCR or the right -hand side of Equa-
and entrance losses. Solution by Crane method: By tion (52) is as follows:
Solution via Excel spreadsheet using the plot-correlation as given in
=SQRT(K*( 1+r)/ (2*(K+2*LN(l / r)))). O
Solve Equations (16) and (17) simulta- Refer ence [4], (y = 1.4, and K = 45) we
neously: get the following critical values after
r = (P2 ! P0)cR = 0.048 interpolation between K =40 and K = Subcritical flow
Hand calculation: (t:J> / Po)cR = 0.952 100: Graphical methods can give a clue how
From Par t 6 of spreadsheet : (t:J> I Po)cR = 0.885 to deal with subcritical flow. The sub-
YcR = 0.68 YcR = 0.72 critical values for Y can be determined
Critical pressure, P2 = r X P0 = 0.048 The critical pressure, P2 = 759 kPaa by linearly interpolating between (Y =
X 6,600 = 317 kPaa Since P2 > PA, the pipe will choke. 1, t:J> I P0 = 0) and (YCR, [t:J> I Pol CR). If
Since P2 > PA, the pipe will choke. Therefore, for flow estimation, t:J> will Y and [t:J> I PolCR are plotted on y and x
Therefore, for flow estimation, t:J> will be (6,600 - 759) = 5,841 kPa axis respectively, a linear equation to
be (6,600 - 317) = 6,283 kPa. Use Equation (49) with these data . determine intermediate values for Y
Using Equation (49), critical mass Critical mass flow, W = 19,130 kg/h can be written as Equation (18).
flow, W = 18,724 kg/h The plot-based correlation presents
Note that we alternatively can use limitations, as the data wer e outside
(18)
Equation (48) to estimate the critical the graphed range. K had to be ex-
mass flow, by applying the calcu- trapolated and only "I= 1.3 and 1.4 are Solve for slope:
lated value of r directly. Th e for - plotted. The extrapolation process, 1- YcR
mula-based correlation provides an and referring the values from plots,
opportunity to input actual values of are both time consuming. m = ( ll?jpJCR (19)

66 CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002

 Engineering Practice

Solved via spreadsheet. The prob- (M'IPo)cR = 0.952). Interpolating linearly between
lem is the same one as above with one Subcritical Y= 0.816 (Y=l , M' IP0 = 0) and (YcR = 0.72,
change. The downstream pressure is Use Equation (49) for flow. M' I Po)cR = 0.545), or read directly
3,000 kPaa and not 200 kPaa, which is W = 17,028 kg/h Reference [4),
a significant difference. Solved by Crane method: Subcritical Y = 0.82
By using the plot-correlation in Refer- Using Equation (49) for flowrate.
r = (P2 1P0 )cR = 0.048
ence [4), (y =1.4, and K =45) we get the The mass flow, W = 17,104 kg/h. This
(M'IPo)cR = 0.952
following critical values after interpo- is a close check. Obviously, the graph-
YcR = 0.68
lation between K =40 and K =100: ical correlation deviates less in sub-
The critical pressure, P2 = 317 kPaa (M'IPo)cR = 0.885 critical applications. •
Since P2 < PA, the pipe will not YCR = 0.72 Edited by Peter M. Silverberg
choke. Therefore, M' will be (6,600 - The critical pressure, P2 = 759 kPaa
3,000) kPaa. = 3,600 kPa Since P2 < PA, the pipe will not Author
Sub-critical (M' I P o)suB = 3,60016,600 choke. Therefore, M' will be (6,600 - Sunil Kumar is senior
= 0.545 3,000) kPaa = 3,600 kPa process engineer at Worley
Engineering (250 St. Georges
Interpolate linearly between (Y=l, Subcritical (M' IPo)suB = Terrace, Perth, Western
M' IP0 = 0) and (YcR = 0.68, 3,60016,600 = 0.545 Australia 6000; Phone: 61-
8-92788111; Fax: 62-21-
5746709; Email: sunil.kumar
References 4. Crane Co., op cit., p. A-22, . @worley.com.au). He has 12
5. Smith, J . M., and Van Ness, H. C., '!Intl"O- years process design experi-
1. Crane Co., "Flow of Fluids," Publication ence for a variety of offshore
410M, p. 1- 9, Equation 1-11, Joliet, m., duction to Chemical Engineering Thermody-
namics," 4th ed., Sec. 7.2 and 7.3, McGraw- oil and gas r,roduction plat-
March 1988. forms, petro eum refineries,
Hill, New York.
2. Miller, R. W., "Flow Measurement Engineer- petrochemical plants and gas processing
ing Handbook," 2nd ed., Equation 13.14, Mc- 6. Crane Co., op cu., Section 3, Equations 3-14, 3-15. plants. Before Worley, he was employed by
Graw-Hill, New York. 7. Coulson, J. M., and Richardson, J. F., Bechtel. He has a M.Tech. in chemical engi-
3. Kern, R., How to Size Piping as Gas Expands "Chemical Engineering," Vol. 1, 3rd ed., p. neering from Harcourt Butler Technological Jn.
at Flow Conditions, Chem. Eng., Oct. 13, 1975. 30, Equation 2. 71, Per~amon International stitute. He is a member of the Institute of Engi-
Library, Butterwort Hememann. neers (Australia).

-i••i;tii1•1~••di~••lill
New Filtration Technology Keeps Cooling Tower Water Clean
Typical Typical
Full Side
Flow Stream
filtration filtration
using of basin
existing using a
pump. recirculating
pump.

Typical Typical
Side
Stream
filtration
using a
booster
@ Side
Stream
filtration
using
existing
pump. pump.

www.orival.com • filters@orival.com
very cooling tower has its unique dirt specific gravity, even lighter than water. seconds and does not interrupt flow.
E conditions, space constraints and other Typical examples include: airborne parti- Reclaim units recirculate rinse water mak-
special characteristics. Filtration systems cles, microbiological growth, bugs, sand, ing filtration a zero discharge operation. For
must take these into account in order to scale, algae, rust, etc. more information and system design assistance,
maintain particle free water. Line pressure powered, they permit use of contact: Orival, Automatic Self-Cleaning
Orival filters are versatile enough to the ex1stmg cooling tower pump, Filters, 40 N. Van Brunt St., Engle«KXXI, NJ
meet these criteria. They remove a wide lowering initial system cost. The fully au- 07631 . (201) 568-3311 , (800) 567-9767,
range of dirt down to micron size, of any tomatic self cleaning cycle takes Fax: (201) 568-1916.
CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002 67
 Engineering Practice

unraveling the Mvriad 01

Models Used in Modern
Automation Proiects
Greg M artin, M ark Funderburk and Keeping track of the models and their I/Os will avoid
Michael Sugars
Pavilion Technologies, Inc. confusion and ensure proper implementation
ontinuous process automation

C has evolved in the process indus-

tries from single, model-based
property predictors that support
single-loop control applications, to
multi-unit optimization methods that
Property
predictor
Property
predictor

provides targets to multiple linear and

nonlinear model-predictive control
Process Process
(MPCf applications. Models are the
foundation of these control and opti-
mization solutions. FIGURE 1 . Schematic of a property FIGURE 2. Introducing model-predictive
There may be as many as six differ- predictor with a PID controller control (MPC) to a property predictor
ent types of models in one solution, in-
cluding models in the optimizer, the for one model are necessarily related to analysis. If a neural network model is
controller, the supporting property the I/Os of another model. There is used, a few months of process and labo-
predictors, and the forward chaining often confusion about these various ratory values are required to train the
logic. This multiplicity of models can models in terms of what they do and model. The model is implemented in an
cause confusion in discussion and what are the proper I/Os for a particu- online runtime package whereby it can
even occasionally in application. lar model. Furthermore, since these provide periodic output to the program-
This article describes the different models are implemented as parts of a mable-logic controller (PLC) or distrib-
types of models and their purposes, cascade in the overall application, uted control system (DCS). For the ex-
and discusses how they are imple- some inputs may be outputs of a super- ample shown in Figure 1, the online
mented, and how they relate to the visory application. It is important to prediction provides feedback to a con-
other models in the overall applica- understand where each model fits into ventional proportional-integral-deriv-
tion. References are provided that pro- the overall heirarchy of the cascade. ative (PID) controller.
vide details for the particular models
described. Property predictors Model predictive control
Online correlations for laboratory A property predictor providing feed-
Introduction analyses, based on neural network back to an MPC is shown in Figure 2.
An automation project in the process models, were introduced in 1996 as This is a multivariable MPC applica-
industries may include several types of "soft sensors" [1] . Since then, these tion, which has three manipulated
applications, each containing a model types of models have become known variables (MVs). Each MV is the set-
of the process or a part ofit. The mod- as virtual online analyzers or property point (SP) of a PID controller.
els in a single project may include: predictors. Applications range from The property predictor has a model in
• Property predictors petroleum refining to food processing it, and the MPC has a model in it. These
• MPC [2]. Property predictors have also been models are quite different, however, as
• Nonlinear MPC gains extended to include physical models the property predictor model is a
• Forward chaining for distillation and the switching of steady-state relationship (used to pre-
• Process optimization different crude feedstocks [3, 4]. dict a property) and the MPC model is a
• Unit optimization A property-predictor signal-flow dia- matrix of step-response-dynamic mod-
• Multi-unit optimization gram is shown in Figure 1. Several mea- els (used to predict the CV response).
Each model is characterized by its in- surements from the process are used to The I/Os of these models are re-
puts and outputs (VOs), and the VOs calculate one output, usually a product lated. Inputs to the property predictor
68 CHEMICAL ENGINEERING WWW.CHE.COM OCTOBER 2002