AN ABSTRACT OF THE THESIS OF Marcio Matandos for the degree of Master of Science in Chemical Engineering presented on December 12, 1991 . Title: Use of Orthogonal Collocation in the Dynamic Simulation of Staged Separation Processes Redacted for Privacy Abstract approved: Keith L. Levien Two basic approaches to reduce computational requirements for solving distillation problems have been studied: simplifications of the model based on physical approximations and order reduction techniques based on numerical approximations. Several problems have been studied using full and reduced-order techniques along with the following distillation models: Constant Molar Overflow, Constant Molar Holdup and Time-Dependent Molar Holdup. Steady-state results show excellent agreement in the profiles obtained using orthogonal collocation and demonstrate that with an order reduction of up to 54%, reduced-order models yield better results than physically simpler models. Step responses demonstrate that with a reduction in computing time of the order of 60% the method still provides better dynamic simulations than those obtained using physical simplifications. Frequency response data obtained
117
Embed
Redacted for Privacy - CORE · other parameters are assumed constant within sections of the column. Treybal (1985) and McCabe and Smith (1985) present detailed treatments on graphical
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
AN ABSTRACT OF THE THESIS OF
Marcio Matandos for the degree of Master of Science in Chemical
Engineering presented on December 12, 1991 .
Title: Use of Orthogonal Collocation in the Dynamic Simulation of Staged
Separation Processes
Redacted for PrivacyAbstract approved:
Keith L. Levien
Two basic approaches to reduce computational requirements for solving
distillation problems have been studied: simplifications of the model based on
physical approximations and order reduction techniques based on numerical
approximations.
Several problems have been studied using full and reduced-order
techniques along with the following distillation models: Constant Molar
Overflow, Constant Molar Holdup and Time-Dependent Molar Holdup.
Steady-state results show excellent agreement in the profiles obtained using
orthogonal collocation and demonstrate that with an order reduction of up to
54%, reduced-order models yield better results than physically simpler models.
Step responses demonstrate that with a reduction in computing time of the
order of 60% the method still provides better dynamic simulations than those
obtained using physical simplifications. Frequency response data obtained
from pulse tests has been used to verify that reduced-order solutions preserve
the dynamic characteristics of the original full-order system while physical
simplifications do not.
The orthogonal collocation technique is also applied to a coupled columns
scheme with good results.
Use of Orthogonal Collocation in theDynamic Simulation of Staged Separation Processes
by
Marcio Matandos
A THESIS
submitted to
Oregon State University
in partial fulfillment ofthe requirements for the
degree of
Master of Science
Completed December 12, 1991
Commencement Tune 1992
APPROVED:
Redacted for PrivacyAssistant Professor of Chemical Engineering in charge of major
rm i n
Redacted for PrivacyHead of Depa4t#nent of Chemical Engineering
Redacted for PrivacyDean of Gr....6.r
Date thesis is presented December 12, 1991
Typed by Marcio Matandos for Marcio Matandos
ACKNOWLEDGEMENTS
It is hard to imagine that in a personal project such as this one, so many
inevitably become involved and contribute in so many ways.
Among those to whom I feel particularly indebted are Dr. Keith Levien,
whose trust, enthusiasm and guidance throughout the entire course of this
work have been decisive in accomplishing all the goals that were initially set. I
gratefully acknowledge the Teaching Assistantship provided by our Chemical
Engineering Department on Fall 1990 (it was quite an experience!) and thank
very much all faculty members for their dedication and friendship.
To all our friends in Corvallis, thanks a whole lot for making all those
days something to be missed forever.
To my parents, Nicolas and Maria Matandos, and parents-in-law,
Valentino and Oddete Chies, for their love and support throughout the entire
course of this work, my deepest appreciation.
My very special thanks to my son, Bruno, for bringing a new meaning
to even the most trivial things in life and to his mommy, Maria, for her love,
care and patience. Thank you VERY MUCH for being so nice to be with. The
cooperation I received from you guys was simply out of this world!!! I'm
afraid words would never be enough to express all my gratitude to both of
you.
TABLE OF CONTENTS
1-INTRODUCTION 1
1.1-Terminology 3
1.2-Approach to the Problem 41.2.1-Equilibrium-Staged Separation Operations 41.2.2-Systems of Differential and Algebraic Equations 81.2.3-Orthogonal Collocation 9
1.3-Literature Survey 13
2-DEVELOPMENT OF THE MODELS 18
2.1-General Assumptions 18
2.2-Full-Order Models 192.2.1-The Constant Molar Overflow (CMO) Model 252.2.2-The Constant Molar Holdup (CMH) Model 282.2.3-The Time-Dependent Molar Holdup (TDMH) Model 32
3.1.2.1-Bubble Point Temperatures 453.1.2.2-Fraction of Feed Vaporized 46
3.1.3-Liquid Densities 47
3.2-Collocation Points 48
3.3-Software for Initial Value Problems 49
4-EXAMPLE PROBLEMS 51
4.1-Steady-State Results 51
4.1.1-Temperature Profiles 574.1.2-Composition Profiles 574..3-Liquid and Vapor Flowrate Profiles 64
4.2-Step Tests 66
4.3-Pulse Tests 76
4.4-Coupled Columns 78
4.5-Conclusions 93
NOTATION 94
BIBLIOGRAPHY 97
APPENDICES
A. Sequence of Computations 99
B. The Polynomial Interpolation Technique 103
LIST OF FIGURES
Figure Page
1.1 - Schematic diagram and notation for an equilibrium stage. 5
1.2 - Schematic diagram illustrating thecontinuous position variable s withina separation module. 10
2.1 - Schematic diagram and notation for a distillation column. 20
2.2 - Schematic diagram and notation for equilibrium stage 21
2.3 - Schematic diagram and notation for condenser. 22
2.4 - Schematic diagram and notation for reboiler. 23
2.5 - Mass and energy balance envelopes utilized in thederivation of the equations for calculating liquid andvapor flowrates. 31
2.6 - Geometry utilized for calculating liquid flowratesleaving an equilibrium stage using the Francis weirformula. 33
2.7 - Schematic diagram and notation utilized when theorder-reduction technique is applied to a distillationcolumn. 37
4.1 Steady-state temperature profiles for exampleproblem in Sec.4.1 58
4.2 - Steady-state C1 composition profiles for exampleproblem in Sec.4.1. 59
4.3 Steady-state C2 composition profiles for exampleproblem in Sec.4.1. 60
4.4 - Steady-state C3 composition profiles for exampleproblem in Sec.4.1. 61
4.5 - Steady-state C4 composition profiles for exampleproblem in Sec.4.1. 62
Figure Page
4.6 - Steady-state C5 composition profiles for exampleproblem in Sec.4.1. 63
4.7 Steady-state liquid and vapor flowrate profiles for exampleproblem in Sec.4.1. 65
4.8 - Initial and final temperature profiles for example problemin Sec.4.2. 68
4.9 - Step responses for distillate temperature for exampleproblem in Sec.4 2 69
4.10 - Step responses for C2 distillate composition for exampleproblem in Sec.4.2 70
4.11 - Step responses for C3 distillate composition for exampleproblem in Sec.4.2 71
4.12 Step responses for C4 distillate composition for exampleproblem in Sec.4 2 72
4.13 - Initial and final molar holdup profiles for exampleproblem in Sec.4.2 75
4.14 - Pulse responses for C2 distillate composition for exampleproblem in Sec 4 3 77
4.15 - Bode diagram for full-order pulse test responses in Sec.4.3. 79
4.16 - Bode diagram for the TDMH model pulse test responsesin Sec.4 3 80
4.17 - Bode diagram for the CMH model pulse test responsesin Sec.4.3 81
4.18 - Bode diagram for the CMO model pulse test responsesin Sec.4.3 82
4.19 - Specifications for the coupled columns scheme utilizedin Sec.4.4 84
4.20 - Schematic diagram utilized when the order reduction techniqueis applied to the coupled columns problem in Sec.4.4. 85
Figure
4.21 - Steady-state temperature profiles for theexample problem discussed in Sec 4 4
Page
87
4.22 - Steady-state C2 liquid composition profilesfor the example problem discussed in Sec.4.4. 88
4.23 - Steady-state C3 liquid composition profilesfor the example problem discussed in Sec.4.4. 89
4.24 - Steady-state C4 liquid composition profilesfor the example problem discussed in Sec.4.4. 90
4.25 - Step responses for C2 distillate compositionfor the example problem discussed in Sec.4.4. 91
4.26 - Step responses for C3 distillate compositionfor the example problem discussed in Sec.4.4. 92
LIST OF APPENDIX FIGURES
Figure Page
A.1 - Flowsheet diagram demonstrating the sequenceof computations utilized for solving a distillationproblem using the CMO model. 100
A.2 - Flowsheet diagram demonstrating the sequenceof computations utilized for solving a distillationproblem using the CMH model. 101
A.3 - Flowsheet diagram demonstrating the sequenceof computations utilized for solving a distillationproblem using the TDMH model. 102
B.1 - C2 steady-state composition profile for the rectifyingmodule in the column treated in Section 4.1 obtainedusing the lx1 reduced-order TDMH model. 104
LIST OF TABLES
Table Page
2-1 Design variables defining the distillation models on Sec 2 2 27
3-1 - Coefficients for liquid and vapor enthalpies 44
3-2 Coefficients for equilibrium constants. 45
3-3 - Critical constants for liquid densities. 47
4-1 - Specifications for the example problems discussed in Chapter 4. 52
4-2 - Collocation points utilized for determining steady-stateconditions for the example problem of Sec.4.1 usingthe orthogonal collocation technique. 53
4-3 - Steady-state conditions for distillate and bottomsproducts for example problem discussed in Sec.4.1. 55
4-4 - Mean squared errors for steady-stateprofiles of example problem in Sec.4.1. 56
4-5 - Collocation points utilized in the solutionof the example problem discussed in Sec.4.2. 67
4-6 - Relative computing times and integral absolute errorsfor distillate product in example problem of Sec 4 2 73
4-7 - Collocation points utilized in the solution ofthe example problem discussed in Sec.4.4. 86
USE OF ORTHOGONAL COLLOCATION IN THEDYNAMIC SIMULATION OF STAGED SEPARATION PROCESSES
1- INTRODUCTION
Rigorous analytical representation of equilibrium-staged separation
operations includes a large number of nonlinear mass, energy and equilibrium
relationships which must be simultaneously satisfied at each stage. Therefore, a
significant computational effort is usually required in order to solve
multistage/multicomponent separation problems. One approach to reduce
computational requirements is to somehow reduce the number of equations
involved. Two alternatives have been used: either simplify the model based on
assumptions about physical properties and material and energy relations or
simplify the solution method by solving the stage equations only at certain
locations (collocation points).
The first approach has been extensively applied to graphical and shortcut
multistage calculations, where pressure, molar overflow, relative volatility, and
other parameters are assumed constant within sections of the column. Treybal
(1985) and McCabe and Smith (1985) present detailed treatments on graphical
multistage calculations. Henley and Seader (1981) and Wankat (1988) discuss
approximate analytical methods applied to steady-state calculations. It will be
demonstrated that even though the computing effort can be significantly
reduced, such physical simplifications can lead to substantial errors and
2
therefore are suitable only for preliminary analysis and design of distillation
columns.
The second approach, however, will be shown to yield surprisingly
accurate yet computationally cheap results. The models based on this approach
are called reduced-order models. This work attempts to demonstrate the
advantages of one method of reduced-order modeling, orthogonal collocation
(Stewart et al., 1985), for the dynamic simulation of distillation columns based
on more detailed physical models than previously used. To accomplish this,
the following physical models have been treated:
Model # 1 - Constant Molar Overflow (CMO)
Model # 2 - Constant Molar Holdup (CMH)
Model # 3 - Time-Dependent Molar Holdup (TDMH)
The approach utilized to achieve the goals proposed in this thesis consists
of assuming that the full-order TDMH model above yields "true" dynamic
responses and steady-state results. Several distillation problems have been
studied using full and reduced-order techniques for each of the three physical
models. The following aspects have been compared against the full-order
TDMH model:
Steady-state profiles for temperature and composition within a column
3
have been used to verify the accuracy of each method in reproducing
steady-state profiles obtained by the "true" model.
Dynamic responses of column products to step changes of input variables
have been used to verify the ability of each method to predict column
dynamics and to compare the computing time required to solve the
problem.
Dynamic pulse tests have been simulated and analyzed to generate the
frequency response data necessary for determining whether the dynamic
characteristics of the "true" model are well represented.
In addition, in order to demonstrate the suitability of the method for
treating more complex separation systems, orthogonal collocation is applied to
the dynamic simulation of a coupled-columns scheme.
1.1-Terminology
The following definitions are used in this thesis:
Full-order model refers to a physical model where all pertinent
equations are applied at each stage.
4
Reduced-order model refers to a simplification of the physical model
where the number of locations at which the equations are applied has
been reduced by use of numerical techniques.
Section refers to a physical compartment within a column where a
specific separation operation is taking place.
Module represents a set of (not necessarily physical) equilibrium
stages placed one atop the other where no sidestreams are introduced
nor drawn.
Computing time represents the time (expressed in CPU seconds)
required to solve a simulation and depends on the accuracy criteria and
machine being used.
Computing effort refers to the overall amount of mathematical
operations required to solve a simulation problem.
Table 4-3 - Steady-state conditions for distillate and bottoms products for example problemdiscussed in Sec.4.1(T - deg.F, H&S - results obtained by Henley and Seader).
L (mol/h)2 22.9425 121.6697 1.6692E-7 22.9886 147.0180 22229. 0.0 0.0
V (mol/h)2 22.7371 124.6475 3.5685E-6 22.7668 140.3544 22229. 0.0 0.0
Table 4-4 - Mean squared errors for steady-state profiles of example problem in Sec.4.1.Note: Errors computed for full-order models are relative to the TDMH model and errors computed
for reduced-order models are relative to the base model.
57
4.1.1-Temperature Profiles
Inspection of profiles in Fig.4.1(a), and respective errors from Table
4-4 indicate that TDMH and CMH models yield essentially the same
steady-state profiles since at steady-state, mass and energy requirements
are satisfied by both these models. However, energy effects are not
accounted for in the CMO model, and the temperature profile obtained
is noticeably different (MSE=17.1679). Errors in Table 4-4 indicate that
even lx1 collocation approximation applied to the CMH model yields
smaller errors (MSE=15.4667) than the full-order CMO model. Errors
obtained by reduced-order methods applied to either TDMH, CMH or
CMO models have the same order of magnitude demonstrating that the
method yields acceptable results regardless of the physical model being
used.
4.1.2-Composition Profiles
Composition profiles for components C1 through C5 are shown in
Figs.4.2, 4.3, 4.4, 4.5 and 4.6 respectively. Inspection of Fig.4.2 and Table
4-4 shows that when a steep profile exists, such as for component C1, 1
and 2 point collocation techniques are unable to reasonably predict the
composition in the rectifying section and negative mole fractions may
180
150
120
90
60
30
0
180
150
oa.8 120
90
60
30
38
A
0Ei
0 0 0 0 TDMH00000 CMH00000 cmo
0I I I I 1 1 I 1 1 1 1
1 2 3 4 5 6 7 8 9 10 11
STAGE NUMBER
(a)
12
150
cit.8 120
90
tq] 60a.
180
150
120
90
60
30
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 B 9 10 11 12
STAGE NUMBER
(c)
STAGE NUMBER
(d)
Figure 4.1 - Steady-state temperature profiles for example problem in Sec.4.1. (a) - Full-order models,(bJ - TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) CMO full and reduced-order. oo
ui
0.10
0.09 - - 0 0 0 TDMH0.080.07
00000 CMH[moon CMO
0.06
0.05
0.04
0.03 U
0.020 a it
0.01
0.00
0'1
I
2'3
l
4l 1 T ? I I?5 6 7 8 9 10 11 12
STAGE NUMBER
(a)0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
z
0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0.10
0.09
o 0.08
y 0.07
IA- 0.06
Ili 0.0502 0.04o5 0.03a=, 0.02
0.01
0.00
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
STAGE NUMBER
(c)
STAGE NUMBER
(d)
Figure 4.2 - Steady-state C1-composition profiles for example problem in Sec.4.1. (a) Full-order models,(13)-- TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) - CMO full and reduced-order. ul.0
1.0
0.9
o 0.8
5 0.7
0.6
I 0.50x 0.4
nu..,/
E. 0.2
0.1
0.0
0 0 0 TDMH00000 CMH
O Docioo CMO
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(a)
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
STAGE NUMBER STAGE NUMBER
(c) (d)
Figure 4.3 Steady-state C2-composition profiles for example problem in Sec.4.1. (a) - Full-order models,(1:4 - TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) - CMO full and reduced-order.
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
1.0
0.9 -0.8 -0.7
0.6 ---,
0.5 - o
ki ilo li@
W
a
0.4 - @
0.30
0.2 -0.1 -
19
@o
a.. 0 0 0 TDMH00000 CMH0000 CMO
0.0 111111[1 IIII0 1 2 3 4 5 6 7 8 9 10 11 12
STAGE NUMBER
(a)1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(c)
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
0 1 2 3 4 5 6 7 8 9 10 11 12
STAGE NUMBER
(d)
Figure 4.4 Steady-state C3-composition profiles for example problem in Sec.4.1. (a) Full-order models,(1:9-- TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) - CMO full and reduced-order. cn
--+
0.10
0.09
0.080.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0.10
0.09
o 0.080.07
Li- 0.06
ti 0.0502 0.04
5 0.030.02
0.01
0.00
0
0 0 0 0 TDMH00000 CMH0000 CM()
111111 2 3 4 5 6 7
STAGE NUMBER
(a)
8I I I
9 10 11 12
0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00
0.10
0.09
o 0.08g 0.07E 0.06t2, 0.050
0.04
0.03
0.02
0.01
0.00
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
STAGE NUMBER STAGE NUMBER
(c) (d)
Figure 4.5 - Steady-state C4-composition profiles for example problem in Sec.4.1. (a) - Full-order models,(b) - TDMH full and reduced-order, (c) CMH full and reduced-order, (d) - CMO full and reduced-order.
rs.)
0.10
0.09 0 0 0 0 0 TOMH
0.08 00000 CMHEmoo CMO
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00I I I lit I
0.10
0.09zo 0.08
0.07
6- 0.060.050
m 0.04o5 0.03o=, 0.02
0.01
0.00
0 1 2 3 4 5 6 7 8 9 10 11
STAGE NUMBER
(a)
12
0.100.09
0.08
0.07
0.06
0.050.04
0.03
0.02
0.01
0.00
0.10
0.09
0.080.07
0.06
0.050.04
0.03
0.02
0.01
0.00
4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(b)
STAGE NUMBER
(c)
STAGE NUMBER
(d)
Figure 4.6 - Steady-state C5-composition profiles for example problem in Sec.4.1. (a) - Full-order models,(133- - TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) CMO full and reduced-order. wa.
64
arise (as seen in Figs.4.2(b), (c) and (d)). This agrees with Srivastava and
Joseph's (1985a) observations and consequently for this case, the full-
order CMO model yields smaller errors than those obtained by either
reduced-order method. Nevertheless, these reduced-order profiles are
useful in illustrating the polynomial interpolation where lx1 and 2x2
collocation profiles are represented by straight lines and parabolas (first
and second-order approximations) respectively.
Profiles and resulting errors obtained for remaining components, C2
through C5, using the order reduction technique applied to either
TDMH or CMH models once again indicate that even the lower order
1x1 approximation yields better profiles than the full-order CMO model.
However, since the compositions of non-key component C4 in the
rectifying section are low and form a relatively steep profile, the
solution obtained using orthogonal collocation yields negative values for
stages close to the top of the column. This could be avoided using a
higher order approximation, e.g. 3x2.
4.1.3-Liquid and Vapor Flowrate Profiles
Figure 4.7 presents liquid and vapor flowrate profiles obtained by
each method of solution utilized. Comparison of full-order profiles for
the three physical models show that constant flowrates from the CMO
2000 2000
.--.1500 .E 1500..
Zo -6
E Liquid E
F 1000 .iii 1000
& &3 30 0iL..,
_t500 1,_ 500
4 5 6 7 8 9 10 11 12
STAGE NUMBER
(a)
0 11111IIIIII0 1 2 3 4 5 6 7 8 9
STAGE NUMBER
(C)
10 11 12
0
2000
0111111111111 2 3 4 5 6 7 8
STAGE NUMBER
(b)
9 10 11 12
2 1500O
1.7.W 1000
0500
Vaporo*
Liquid
oo 000 CMO00000 2*2
1*1
0
1 I I I I I I I I 1 I
1 2 3 4 5 6 7 8
STAGE NUMBER
(d)
9 10 11 12
Figure 4.7 - Steady-state liquid and vapor flowrate_profiles for example problem in Sec.4.1. (a) - Full-orderUmodels, (b) TDMH full and reduced-order, (c) - .4H full and reduced, (d) - CMO full and reduced. cr.
66
model are not very accurate (MSE=22229). However, profiles obtained
by the reduced-order technique for TDMH and CMH models are
noticeably better.
4.2-Step Tests
Example problem 2 from Table 4-1 (devised for illustration purposes in this
work) is utilized to demonstrate the reduction in computing time and accuracy
associated with the use of reduced-order techniques. This problem deals with
the transient response of a column due to the introduction of a step decrease
in the distillate flowrate (from 300 to 250 mol/h) and consists of determining
distillate conditions and computing time spent in simulating 60 process hours.
Since the reboiler duty is maintained constant, the decrease in distillate
flowrate is equivalent to an increase in the reflux ratio.
Collocation points utilized for solving the problem using the order
reduction technique are presented in Table 4-5 and provide a reduction in the
number of equations from 60 (78 for the TDMH model) to 24 (30), 27 (35) and
30 (38) corresponding to an order reduction of approximately 60, 55 and 50%
respectively. Since CMH and CMO models use constant stage molar holdups,
these values were calculated as the average initial steady-state stage holdups
from the corresponding TDMH model.
Figure 4.8 shows the change of the column temperature profile caused by
67
the introduction of the step in the distillate flowrate, where a decrease in
distillate temperature results from the increase in purity of the lighter
component C2. This is as expected since a decrease in distillate flowrate
corresponds to an increase in reflux ratio. Since the effect of the step on
bottoms product temperature is less, only the distillate product response will
Table 4-5 - Collocation points utilized in thesolution of the example problem discussed in Sec.4.2
Dynamic responses obtained for distillate temperature and composition are
shown in Figures 4.9, 4.10, 4.11 and 4.12. Table 4-6 presents the computing
times (relative to the full-order TDMH model) required to determine these
responses (on a 386DX/33MHz machine using FORTRAN coding compiled
using the Microsoft FORTRAN 4.1 compiler) and respective errors, calculated
200
.. 1606.
CSw0Weg 120
igo.zW1
80
40
10 15 20STAGE NUMBER
Figure 4.8 - Initial and final temperature profiles forexample problem in Sec.4.2. (Solid lines represent
initial steady-state conditions and dashed linesrepresent conditions after 60 process hours).
68
c: 90
00
65 ; ----
40
C 900
TDMHCMHCM0
111i1j1111110 20 30 40 50
TIME (hours)
(a)
60
LaetD .
ii.., 65 ...a. .....2WI--
CMH
5 - -- 4+2-.1 3+2
2.28 40 f
1
11 ' I ' 1 I I 1
11
O 10 20 30 40 50 60TIME (hours)
C 90oI
V
:011 6o.
52
40
...
TDMH4+23+22.2
1 I 1
0 10 20 30 40TIME (hours)
C 906-o
ik 65a.2
5
8 40
(b)
50 60
CM04«23+22+2
....1 I I I 1
10 20 30 40TIME (hours)
1
50
(c) (d)
Fligure 4.9 - Step responses for distillate temperature for example problem in Sec.4.2. (a) Full-order models, (b) -WMH full and reduced-order, (c) - CMH full and reduced-order, (d) - CM0 full and reduced-order.
60
1.00
0.90
........
.."-''''
...................................
OP
- ..../
..
.- .0.80 '
TDMH, CMHCM0
0.70 T1 II II I I I
10 20 30 40 50 60TIME (hours)
(a)1.00
0.90 1
0.80 CMH4.23*22*2
0.70 Ii
1 II 11 T
10 20 30 40 50 60TIME (hours)
(c)
1.00
0.90
0.80
0.70
1.00
0.90
0.80
0.70
...................................
TDMH4*23*22*2
I I I I
0 10 20 30 40TIME (hours)
(b)
I
50 60
....................................r--,
CM04*23.22*2
I 1 I i I 1 I
0 10 20 30 40TIME (hours)
(d)
i
50i
Figure 4.10 - Step responses for C2 distillate composition for example problem in Sec.4.2. (a)-Full-order models,(b)--TDMH full and reduced-order, (c)-CMH full and reduced-order, (41)-CM0 full and reduced order.
60
0.25
0.20
0.15;
0.10
0.05 1
0.00 ( Iv
I i ' I ' I ' I I
0 10 20 30 40TIME (hours)
(a)
TDMHCMHCM0
0.25
0.20
0.15
0.10
0.05
0.00
50 60
10 20 30 40 50 60TIME (hours)
(c)
0.25
0.20
0.15
0.10
TDM H4+23+2
0.05
2+20.00
1 I I
0 10 20 30 40 50 60TIME (hours)
(b)0.25
Cm04+20.203+22+2
0.15
0.10
0.05 ...................... 4.7WM.
0.00I I r--I 1 1 1
11
1
10 20 30 40 50 60TIME (hours)
(d)
Figure 4.11 - Step responses for C3 distillate composition in example problem in Sec.4.2. (a)-Full-order models,(b)--TDMH full and reduced-order, (c)-CMH full and reduced-order, (d) -CMO full and reduced-order.
0.015
0.005
TDMHCMHCM0
0.0050 10 20 30 40 50
TIME (hours)
(a)
60
0.015
0.005
CMH4*23*22*2
0.0050 10 20 30 40
TIME (hours)
(c)
50 60
0.015
0.005
TDMH4*23*22*2
0.0050 10 20 30 40 50
TIME (hours)
(b)
60
0.015
0.005
CM04*23*22*2
.....
0.0050 10 20 30 40
TIME (hours)
(d)
50
Figure 4.12 - Step responses for C4 distillate composition in example problem in Sec.4.2. (a)-Full-order models,(b TDMH full and reduced-order, (c)-CMH full and reduced-order, (1:1)-CM0 full and reduced-order.
Table 4-6 - Relative computing times and integral absolute errors for distillate product in example problemof Sec.4.2. Note: Computing time for full-order TDMH is 15'30". Errors computed for full-order models
are relative to the TDMH model and errors for reduced-order models are relative to the base model.
74
using an Integral Absolute error norm (IAE) given by
t=60
IAE = f jA(t)-A(t) I dtt=o
where A(t) and A(t) represent respectively the "true" and approximated step
responses for the variable under consideration.
Figure 4.9 demonstrates the dynamic response for the distillate
product temperature. As in Sec.4.1, TDMH and CMO models yield identical
initial and final steady-state values. However, according to Fig.4.9(a), the
speeds of response for each model are clearly different. Comparison of the
IAE's for temperature responses in Table 4-6 for full-order CMH and CMO
models with those using 4x2 TDMH demonstrates the advantages of the
orthogonal collocation technique where, with an order reduction of 50%, the
error obtained by the reduced-order solution (IAE=30.48) is much smaller than
that from either full-order CMH or CMO models (IAE=73.45 and 990.04
respectively). Errors in the CMH and CMO models can be explained by
inspection of Figure 4.13, where initial and final molar holdup profiles along
the column are presented. Since both CMH and CMO models use the constant
molar holdup assumption, variations in molar holdup that occur after the
introduction of the step in the distillate flowrate are ignored causing the
differences observed in the dynamic behavior of the simpler model.
IrIT rill-Ili5 10 15
STAGE NUMBER
20
Figure 4.13 Initial and final molar holdup profilesfor example problem in Sec.4.2. (Solid lines represent
initial steady-state conditions and dashed lines representconditions after 60 process hours).
75
76
4.3-Pulse Tests
In this section, frequency response analysis is utilized to determine
whether reduced-order models preserve the dynamic characteristics of the
original full-order system. To do so, pulse tests have been performed using the
distillation column treated in example problem 2.
Pulse responses for the distillate mole fraction of component C2 are shown
in Figure 4.14 using a deviation variable as follows
YD,;(t) = YD,;(t) -YD,c(to)
These responses have been obtained by starting from the same initial steady-
state conditions as in the previous section. Then a rectangular pulse (from 300
down to 285) is introduced in the distillate flowrate and sustained for 0.215
hours after which the distillate flowrate is returned to its initial value.
Pulse response input/output data carry important information regarding
the dynamic characteristics of the system and can be numerically analyzed to
generate frequency response Bode diagrams for the system. These diagrams
are used in the analysis of linear characteristics and design of linear
controllers. Even though the distillation models are nonlinear, the design of
linear controllers for distillation columns is usually based on linear dynamic
models. For the purpose of controller design, if the collocation solutions result
in the same Bode diagrams as the full-order solution, then the collocation
technique can be used for controller design.
0.003
0.002
0.001
0.000
0.003
0.002
0.001
TDMHCMHCMO
10 15 20TIME (hrs.)
(a)
25 30
0.003
0.002
0.001
0.000
0.003
0.002
0.001
0.000 0.00010 15 20 25 30 0 5 10 15 20 25 30
TDMH4.23«22.2
IIIIIII I tr-5 10 15 20 25 30
TIME (hrs.)
(b)
TIME (hrs.)
(c)
TIME (hrs.)
(d)
Figure 4.14 Pulse responses for C2 distillate composition for example problem in Sec.4.3. (a) Full-order models,(13J - TDMH full and reduced-order, (c) - CMH full and reduced-order, (d) CMO full and reduced-order.
78
The Bode diagram for full-order pulse responses is presented in
Figure 4.15(a) where at higher frequencies, the amplitude ratio (AR) follows a
straight line with slope -1 and the phase angle (PA) asymptotically
approaches -90° indicating first-order dynamics for the systems under
investigation. Model comparisons are based on residual amplitude ratio
(defined as the ratio between the AR of the simplified and true models) and
residual phase angle (defined as the difference between the PA of the
simplified and true models) which, for perfect approximation of the "true"
solution, should equal 1.0 (dimensionless) and 0.0 degrees respectively
throughout the range of frequencies analyzed. In Figure 4.15(b), the residual
AR and PA demonstrate that full-order TDMH, CMH and CMO models differ
with regard to their dynamic characteristics. Inspection of Figures 4.16(a) and
(b) on the other hand, demonstrate that reduced-order techniques applied to
the TDMH model preserve the dynamic characteristics of the full-order system
with remarkable fidelity, yielding even better results with the lower order 2x2
approximation than those obtained using full-order solutions for CMH or
CMO models.
4.4-Coupled Columns
In order to demonstrate the versatility of orthogonal collocation models in
simulating more complex separation systems, its modular structure has been
o00
0 -=00=
m.111::: ::::====Z...:...11. a .1601Ismomilot om......:1;;ammo milMANNINOMilii
Figure 4.15 - (a) Bode diagram for full-order pulse test responses in Sec.4.3. (b) Residual dynamics resultingfrom the physical simplification of the model.
Figure 4.16 - (a) Bode diagram for the TDMH model pulse test responses in Sec.4.3. (b) - Residual dynamicsresulting from the collocation approximation. co
00
OO0
080
CMH4*23*22*2
0.001
0
ta
ZO
O
O
I 1111111
0.01 0.1
FREQUENCY
I
1
(rod/hr.)
10
0.001 0 01 0 1 1
FREQUENCY (rod/hr.)
(a)
10
0
1..411 %EP .1
0.001 0 01 0.1 1
FREQUENCY (rod/hr.)
J
a. 0
0
0
0.001
10
0 01 0 1 1
FREQUENCY (rod/hr.)
(b)
10
Figure 4.17 - (a) Bode diagram for the CMH model pulse test responses in Sec.4.3. (b) - Residual dynamicsresulting from the collocation approximation. co
0=liquid entry point 4.0000 2.81654=reboiler # 2 4.0000
Table 4-7 - Collocation points utilized in thesolution of the example problem discussed in Sec.4.4.
I-
300
250 -
50
0
6 200 -
LAJ
2 100 -w -
REBOILED STRIPPER
MAIN COLUMN
woo FULL ORDER00000 2 POINT COLLOCATION00000 1 POINT COLLOCATION
0I t 1 I r I II-III-1-i
3 6 9 12
STAGE NUMBER
1 I
15 18
Figure 4.21 - Steady-state temperature profiles for theexample problem discussed in Sec.4.4.
87
88
1.0
0.9
0.8
0.7z0i= 0.6
(4. 0.5
0 0.420.3
0.2 -
0.1 -
0.00
co. FULLORDER00000 2 POINT COLLOCATION00000 1 POINT COLLOCATION
MAIN COLUMN
REBOILED STRIPPER
I I r I r I [I ri Iri3 6 9 12
STAGE NUMBER
15 18
Figure 4.22 - Steady-state C2 liquid composition profilesfor the example problem discussed in Sec.4.4.
1.0 -
0.9
0.8
0.7 --
0.6 -
0.5
0.4 -
0.3 -
0.2 -
0.1
0.0
REBOILED STRIPPER
MAIN COLUMN
FULLORDER00000 2 POINT COLLOCATION0000c 1 POINT COLLOCATIONuTilril
0 3 6ITI I I I
9 12
STAGE NUMBER
I
15 18
Figure 4.23 - Steady-state C3 liquid composition profilesfor the example problem discussed in Sec.4.4.
89
1.0
0.9
0.8
0.7z0P. 0.6U
0.5
cl 0.420.3
0.2
0.1
0.0
FULLORDER00000 2 POINT COLLOCATION00000 1 POINT COLLOCATION
MAIN COLUMN
REBOILED STRIPPER
0 3I I i I I 1 1 1 1 1 1
6 9 12 15 18
STAGE NUMBER
Figure 4.24 - Steady-state C4 liquid composition profilesfor the example problem discussed in Sec.4.4.
90
91
0.985
zo
:4 0.980
481z
12 18 24 30 36 42 48 54TIME (hours)
60
Figure 4.25 Step response for C2 distillate compositionfor the example problem discussed in Sec.4.4.
92
z0
cdu_
0.030
0.025
0.020Lu
02...... - . .1 _ - . -
- -
0.015
FULL-ORDER2 POINT COLLOCATION1 POINT COLLOCATION
0.010I I 1 1 I 1 I I I
0 6 12 18 24 30 36 42 48 54 60
TIME (hours)
Figure 4.26 - Step response for C3 distillate compositionfor the example problem discussed in Sec.4.4.
93
4.5-Conclusions
The use of three distillation models (CMO, CMH and TDMH) illustrated
that the orthogonal collocation technique is by no means limited to a particular
selection of state variables and thermodynamic or hydraulic relationships.
Steady-state results obtained demonstrated that the method is remarkably
accurate even at lower order approximations. Step and pulse responses served
to demonstrate that orthogonal collocation techniques can be useful in the
dynamic analysis of staged separation processes. Since conventional process
control techniques rely on linear models derived from step or pulse test
information, model parameters (time constants and steady-state gains) can
differ greatly depending on the physical model used: TDMH, CMH or CMO.
Reduced order solutions of the TDMH model are an attractive tool for
controller design and should be useful for on-line dynamic optimization.
The coupled columns scheme discussed in Sec. 4.4 demonstrated that the
method is flexible and robust enough that even complex separation networks
can be greatly simplified and analyzed.
94
NOTATION
Unless otherwise noted in the text, the definitions below express the
intended meaning for the following symbols
A Cross sectional area for column, m2
B Bottoms product flowrate, mol/h
D Distillate product flowrate, mol/h
E1 Vaporization efficiency of species j
F Feed stream flowrate, mol/h
FW Flow parameter in Eq.(2-15)
hB Molar enthalpy of bottoms product, BTU/mol
hi Liquid molar enthalpy on stage i, BTU/mol
Height of vapor-free liquid over the weir onstage i, m
Height of weir, m
HD Molar enthalpy of distillate product, BTU/mol
HF Molar enthalpy of feed stream, BTU/mol
Hi Vapor molar enthalpy on stage i, BTU/mol
Hw Molar enthalpy of withdrawal stream, BTU/mol
K..zj Equilibrium constant for component j on stage i
1(s,t) Generic variable related to the liquid phase
1 Length of weir
95
Li Liquid flowrate leaving stage i, mol/h
m Number of collocation points in a module
M Number of actual stages in a module
Total molar holdup on stage i, mol
nc Total number of components
N Total number of internal stages in a distillation column
P Legendre and Jacobi orthogonal polynomials
Qc Condenser duty, BTU/h
Qi Liquid volumetric flowrate leaving stage i, GPM
QR Reboiler duty, BTU/h
RB Boilup ratio
s Position variable (continuous)
t Time
Temperature on stage i, deg. F
v(s,t) Generic variable related to the vapor phase
Vapor flowrate leaving stage i, mol/h
W Withdrawal stream flowrate, mol/h
Wi(s,t) Lagrange interpolating polynomial for liquid phasevariables
Wrzp(s,t) Lagrange interpolating polynomial for vapor phasevariables
xii Liquid composition of species j on stage i
XB.4 Composition of species j on bottoms product
XF4
Xwi
YD,j
Z.
Composition of species j on feed stream
Composition of species j on withdrawal stream
Vapor composition of species j on stage i
Composition of species j on distillate product
Composition of species j on feed stream
Subscripts
c,j Species j critical property
f Feed stage location
B Bottoms product
D Distillate product
F Feed stream
r Rectifying section
s Stripping section
Superscripts
Greek symbols
a, 13
Pi
Species molar thermodynamic property
Parameters in Hahn polynomials
Molar liquid density on stage i, mol/m3
Fraction of feed vaporized
96
97
BIBLIOGRAPHY
Cho, Y.S. and B. Joseph, Reduced-Order Steady-State and Dynamic Models forSeparation Processes I Development of the Model Reduction Procedure, AIChEJournal, 29, 2, 261 (1983).
Cho, Y.S. and B. Joseph, Reduced-Order Steady-State and Dynamic Models forSeparation Processes - II Application to Nonlinear Multicomponent Systems, AIChEJournal, 29, 2, 270 (1983).
Cho, Y.S. and B. Joseph, Reduced-Order Models for Separation Columns IIIApplication to Columns with Multiple Feeds and Sidestreams, Computers &Chemical Engineering, 8, 2, 81 (1984).
Finlayson, B.A., Nonlinear Analysis in Chemical Engineering, McGraw-Hill BookCo., New York (1980).
Gani, R., C.A. Ruiz and I.T. Cameron, A Generalized Model For DistillationColumns I Model Description and Applications, Computers & ChemicalEngineering, 10, 3, 181 (1986).
Gani, R., C.A. Ruiz and I.T. Cameron, A Generalized Model For DistillationColumns II - Numerical and Computational Aspects, Computers & ChemicalEngineering, 10, 3, 199 (1986).
Henley, E.J. and J.D. Seader, Equilibrium-Stage Separation Operations in ChemicalEngineering, 1st ed., John Wiley & Sons, Inc., New York (1981).
Holland, C.D. and A.I. Liapis, Computer Methods For Solving Dynamic SeparationProblems, 3rd ed., McGraw-Hill Book Company, Inc., New York (1983).
Howard, G.M., Unsteady State Behavior of Multicomponent Distillation Columns,AIChE Journal, 16, 6, 1023, (1970).
McCabe, W.L., J.C. Smith and P. Harriot, Unit Operations of ChemicalEngineering, McGraw-Hill Book Co., New York (1985).
Osborne, A., The Calculation of Unsteady State Multicomponent Distillation UsingPartial Differential Equations, AIChE Journal, 17, 3, 696 (1971).
Petzold, L.R., A Description of DASSL: A Differential / Algebraic System Solver,SAND82-8637, Sandia National Laboratories (1982).
98
Pugkhem, C., Steady-State Simulation of Distillation Columns Using OrthogonalCollocation, M.S. Project, Oregon State University, Corvallis, OR, (1990).
Smith, J.M. and H.C. Van Ness, Introduction to Chemical EngineeringThermodynamics, McGraw-Hill Book Co., New York, (1987)
Srivastava, R.K. and B. Joseph, Reduced-Order Models for Separation Columns IVTreatment of Columns with Multiple Feeds and Sidestreams Via Spline Fitting,Computers & Chemical Engineering, 11, 2, 159 (1987).
Srivastava, R.K. and B. Joseph, Reduced-Order Models for Separation Columns VSelection of Collocation Points, Computers & Chemical Engineering, 9, 6, 601(1985).
Srivastava, R.K. and B. Joseph, Reduced-Order Models for Separation Columns VIColumns with Steep and Flat Composition Profiles, Computers & ChemicalEngineering, 11, 2, 165 (1987).
Stewart, W.E., K.L. Levien and M. Morari, Simulation of Fractionation byOrthogonal Collocation, Chemical Engineering Science, 40, 3,409 (1985).
Swartz, C.L.E. and W.E. Stewart, A Collocation Approach to Distillation ColumnDesign, AIChE Journal, 32, 11, 1832, (1986).
Treybal, R.E., Mass Trasfer Operations, McGraw-Hill Book Co., New York (1981).
Villadsen, J. and M.L. Michelsen, Solution of Differential Equation Models byPolynomial Approximation, 1st ed., Prentice-Hall, Inc., New Jersey (1978).
Wankat, P.C., Separations in Chemical Engineering : Equilibrium Staged Separations,1st ed., Elsevier Science Publishing Co., Inc., New York (1988).
Wilkinson, W.L. and W.D. Armstrong, An Approximate Method of PredictingComposition Response of a Fractionating Column, Chemical Engineering Science, 7,1, (1957).
Wong, K.T. and R. Luus, Model Reduction of High-Order Multistage Systems bythe Method of Orthogonal Collocation, The Canadian Journal of ChemicalEngineering, 58, 382 (1980).
APPENDICES
99
APPENDIX A
Sequence of Computations
Figures A.1, A.2 and A.3 present the basic sequence of computations
utilized by the FORTRAN programs used for solving the example problems
discussed in Chapter 4. Even though these programs have a very similar
structure, different codes have been written for each physical model and for
the coupled columns scheme.
Copies of the source codes for these programs may be obtained by writing
to:
Dr. Keith L. LevienChemical Engineering DepartmentOregon State UniversityGleeson Hall 103Corvallis, OR 97331-2702
(Calculate vapor compositionsfor each stage
Read inputs fromMisfile
(Calculate feed)conditions
(Set liquid composition ateach stage
(Calculate liquid and vaporflowrates
Read change inmanipulated variable
100
YES
Interpolate liquid and vaporcompositions entering
collocation points
( STOP >if NO
Reset ODE's and callintegrating package
YES
YES
NO 1
NO _
YES
Figure A.1 - Flowsheet diagram demonstrating the sequence of computationsutilized for solving a distillation problem using the CMO model.
(Set liquid composition ateach stage
Read inputs fromdatafile
(Calculate feed )conditions
47
'Calculate bubble point temperatures,vapor compositions and liquid and
vapor enthalpies at each stage
(Calculate differential term forenergy balances
YES
[Interpolate liquid and vapor
compositions and enthalpies forstreams entering collocation points
( STOP) NO
Calculate liquid and vapo9flowrates
101
4--
NO11
YES
YES
NO _
YES
Figure A.2 - Flowsheet diagram demonstrating the sequence of computationsutilized for solving a distillation problem using the CMH model.
Read inputs fromdatafile
(Calculate feed )conditions
( )Set liquid composition andmolar holdups at each stage