CH 07

CHAPTER 7
CONVERTING FROM STEADY STATE

TO DYNAMIC SIMULATION
We are now ready to get into the details of converting a steady-state simulation into a
dynamic one. Basically the additional information that must be provided is the physical
dimensions of the various pieces of equipment.
It is important to remember that pumps and control valves have already been installed in
the steady-state simulation. These are not necessary for steady-state simulation, but they are
vital for a realistic dynamic simulation. Providingsufcient pressure dropover a control valve
at design conditions with the valve at some fraction opening (typically 50%) is crucial for
dynamic controllability. If valve pressure drop is too small, changing the valve opening
from 50% to 100% will, in piping systems with other equipment taking pressure drops,
result in only a fairly small increase in owrate. If a valve saturates, controllability is lost.
The propane/isobutane column developed in Chapter 3 is used in this chapter as a
numerical example. The control valves all have pressure drops of 3 atm. The column
has 32 stages, is fed on stage 14, operates at 16.8 atm, and produces distillate and bottoms
products with impurity levels of 2 mol% isobutane and 1 mol% propane, respectively.
7.1 EQUIPMENT SIZING
The dynamic response of a ow system depends on the owrate and the volume. For a
given owrate, the smaller the volume, the faster the transient response.
The procedure for sizing the distillation column shell has already been discussed (in
Chapter 3). The only remaining issues are the sizes of the reux drum and the column
base. A commonly used heuristic is to set these holdups to allow for 5 min of liquid
holdup when the vessel is 50% full, based on the total liquid entering or leaving the
vessel. For the reux drum, this is the sum of the liquid distillate and the reux. For the
column base, it is the liquid entering the reboiler from the bottom tray.
151
Distillation Design and Control Using Aspen
TM
Simulation, By William L. Luyben
Copyright #2006 John Wiley & Sons, Inc.
These volumetric liquid owrates can be found by clicking on Proles under the block
C1 and then opening the Hydraulics page tab. Figure 7.1 shows the window that opens on
which the volumetric liquid owrate for the reux drum (stage 1) is given as 0.1782 m
3
/s.
Scrolling down to the bottom tray (stage 31) gives a volumetric liquid owrate of
0.3438 m
3
/s. The total volume of the reux drum should be 0.1782(60)(10) 106.9 m
3
and that of the column base 0.3438(60)(10) 206.3 m
3
. Assuming a length to diameter
ratio of 2, the diameters and lengths can be calculated:
Volume
pD
2
4
(2D)
The reux drum is 4.08 m in the diameter and 8.16 m in length. The column base (or reboi-
ler or sump) has a diameter of 5.08 m and a length of 10.16 m.
The values are entered by clicking the Dynamics button on the top toolbar (see Fig. 7.2).
If this button is not displayed, click the View button, then Toolbar, and check the
Dynamics box. The window that opens has several page tabs. On the Reux Drum page
tab, the appropriate diameter and length are entered. The same is done on the Sump
page tab. Finally, the Hydraulics page tab is clicked, and the window shown in
Figure 7.3 opens, on which stage numbers (2 through 31) and the column diameter
(5.91 m) are entered. The default values of weir height and tray spacing are 0.05 and
0.6096 m, respectively.
At this point all equipment has been sized. There remain two items to take care of. The
pressure of the feedstream leaving valve V1 must be exactly equal to the pressure on the
stage where it is fed. The pressure on stage 14 is found by looking in Proles
(1,709,839.11 N/m
2
). The outlet pressure of valve V1 is set equal to this value, and the
simulation is run again. The last thing to do is to click the Pressure Checker button that
is just to the right of the Dynamics button on the top toolbar (see Fig. 7.2). If the plumbing
has been correctly specied, the window shown in Figure 7.4 appears. We are ready to go
into Aspen Dynamics.
Figure 7.1 Hydraulics page tab.
152 CONVERTING FROM STEADY STATE TO DYNAMIC SIMULATION
7.2 EXPORTING TO ASPEN DYNAMICS
Aspen Dynamics uses the steady-state information generated in Aspen Plus, but they are
two different programs with different les. The Aspen Plus le is lename.apw, and there
is also a backup le, lename.bkp, that is generated. The latter le can be used to upgrade
to newer versions of Aspen Plus.
The information from Aspen Plus is exported into Aspen Dynamics by generating
two additional les. The rst is a lename.dynf le that is used in Aspen Dynamics and
is modied to incorporate controllers, plots, and other features. The second le is a
Figure 7.2 Dynamics button.
Figure 7.3 Specifying diameter and weir height.
7.2 EXPORTING TO ASPEN DYNAMICS 153
lenamedyn.appdf le that contains all the physical property information to be used in
Aspen Dynamics.
In the propane/isobutane column example, the Aspen Plus les are called
Example1.apw and Example1.bkp. The les generated and used in Aspen Dynamics are
Example1.dynf and Example1dyn.appdf.
The procedure for exporting is to click on File at the top left corner of the Aspen Plus
window and select Export. The window shown in Figure 7.5 opens, and the dropdown
menu is used to select P driven Dyn Simulation, which is the tenth item on the list.
Then the Save button is clicked.
An error appears (Fig. 7.6a), which informs us that the specied pressure drop (0.0068
atm) is too small for the given weir height and vapor rates. Aspen Dynamics calculates tray
pressure drops rigorously, and they change with vapor and liquid rates. We have two
options: increase the specied pressure drop or decrease the weir height. The latter has
no effect on the steady-state solution, so we go back to Dynamics under the C1 block,
select the Hydraulics page tab, and change the weir height from 0.05 to 0.025 m.
Running the program again, pressure checking, and Exporting again cause the
Figure 7.4 Pressure check.
Figure 7.5 Exporting.
window shown in Figure 7.6b to open. These messages just remind us that level controllers
should be installed once we get into Aspen Dynamics and the typical pump curves are used
in the pumps.
The export is successful, and we are ready to go into Aspen Dynamics.
Figure 7.6 (a) Error message; (b) warning message.
7.2 EXPORTING TO ASPEN DYNAMICS 155
7.3 OPENING THE DYNAMIC SIMULATION IN ASPEN DYNAMICS
The easiest way to open the Example1.dynf le is to double-click on the lename in
Windows Explorer. The screen that opens contains several windows (see Fig. 7.7).
The Process Flowsheet Window is where the control structure will be developed. The
Simulation Messages window is where the progress of the simulation is shown and simu-
lation time is displayed. The window in the upper left corner, Exploring, is where various
types of controllers, control signals, and other elements can be found to drop and drag
onto the owsheet.
The very rst thing to do with any newly imported le is to make an initialization run
to make sure that everything is running. At the very top of the screen there is a little
window that says Dynamic. Clicking the arrow to the right opens the dropdown menu
shown in Figure 7.8a. Select Initialization and click the Run button, which is just to the
upper right. If everything is set up correctly, the window shown in Figure 7.8b opens.
The next thing to do is to make sure that the integrator is working correctly. This is done
by changing from Initialization back to Dynamic and clicking the Run button again. The
Simulation Messages window at the bottom of the screen should start displaying advan-
cing simulation time. An example is shown in Figure 7.8c. Note the green block at the
bottom of the screen. If something is wrong, this block will turn red and you will not
be able to run. The run is stopped by clicking the Pause button, which is the second
button to the right of the Run button.
Figure 7.7 Initial screen in Aspen Dynamics.
The initial owsheet has some default controllers already installed. In this simple,
single-column process there is only one default controller, the pressure controller. It is con-
gured to measure condenser pressure and manipulate condenser heat removal. The action
of the controller, the range of the pressure transmitter, the maximum heat removal rate, and
the controller tuning constants are all set up at some nominal values. We will come back
and look at these later. But this indicates the number of items that must be specied for each
newcontroller that we add to the owsheet. At a minimum, four additional controllers must
be added to achieve effective operation of the column:
1. Reux drum-level controller
2. Base-level controller
3. Feedow controller
4. Tray temperature controller
Which of the four remaining manipulated variables selected to control each of these four is
the issue of control structure selection. A variety of control structures will be discussed
here and in subsequent chapters.
We have not included reux in the list above because the default conguration in Aspen
Dynamics is to x the mass owrate of the reux. We will discuss how this setup can be
changed later in this chapter.
Figure 7.8 (a) Selecting initialization; (b) initialization run successful; (c) making Dynamic run.
7.3 OPENING THE DYNAMIC SIMULATION IN ASPEN DYNAMICS 157
7.4 INSTALLING BASIC CONTROLLERS
Let us gothrough the details of installinga level controller onthe base of the column. Goto the
Exploring window, click on Dynamics under Simulation, and select Control Models. A long
list of alternative controller types and dynamic elements is displayed, as shown in Figure 7.9.
We will use several of these extensively: Dead_time, Lag_1, Multipy, PID, and PIDIncr.
For simple level controllers, which should be proportional only, the PID model is used.
Place the cursor on the PID icon, press and hold the left mouse button, and drag the cursor
to the owsheet window. Releasing the mouse button places a circle on the owsheet,
which is labeled B2 (see Fig. 7.10). This can be renamed by clicking the circle, right-
clicking, selecting Rename block, and typing in the desired label LC11.
Now the control signals must be connected. The process variable signal PV is the vari-
able to be controlled. For this level controller, it is the liquid level on the last stage of the
column (stage 32). We go back to the Exploring window and click on Stream Types under
Dynamics, as shown in Figure 7.11.
Figure 7.9 Control models.
Figure 7.10 Controller placed on owsheet.
Figure 7.11 Selecting control signal.
7.4 INSTALLING BASIC CONTROLLERS 159
To place a control signal on the owsheet, we place the cursor over the ControlSignal
icon, click and hold the left mouse button, and move the cursor to the owsheet. A
number of blue arrows show up that indicate where a control signal can be placed.
Placing the cursor on the arrow coming out of the reboiler and releasing the mouse
button opens the window shown in Figure 7.12, on which we can scroll down to stage
32 and select Level.
Clicking OK connects the control signal line to the reboiler and lets us connect the
other end to the controller (see Fig. 7.13a). Clicking on the blue arrow on the left side
of the controller opens the window shown in Figure 7.13b, where we select LC11.PV.
Clicking OK completes the control signal connection between the reboiler and the level
controller.
The next step is to connect a control signal from the controller to the valve V11 on
the bottoms stream. Dragging and dropping another control signal at the arrow exiting
the controller opens the window shown in Figure 7.14a. We select the LC11.OP and
click OK. Then the line is connected to the valve. The nal LC11 loop is shown in
Figure 7.14b.
To set up the controller, double-click on the LC11 icon on the owsheet. The con-
troller faceplate shown in Figure 7.15a opens. Controller faceplates are where we keep
track of what is going on in the dynamic simulation, set controller parameters, switch
from manual to automatic control, change setpoints, and perform other tasks. The very
rst thing to do is to click the Congure button. This opens the window shown in
Figure 7.15b. Then we click on the Initialize Values button at the bottom of this
window. This provides the steady-state values of the base level (6.35 m) and the control
valve opening (Bias 50%).
Figure 7.12 Attaching control signal.
The action of the controller should be Direct (under Controller action in Fig. 7.15b)
because if the level increases, the signal to the valve should increase (PV", OP") to
remove more bottoms. In some columns, base level is controlled by manipulating a
valve in the feed to the column. In that control structure, the base level controller action
should be Reverse.
Since we want proportional-only control, the controller gain is set equal to 2 and the
integral time is set at a very large number (9999 min, as shown in Fig. 7.15b).
Next the Ranges page tab is clicked, which opens the window shown in Figure 7.16.
The default value for the level transmitter range is 012.7 m. The default value of the con-
troller output range is 0100%. Both of these are what we want, so they require no
changes. The Congure window is then closed.
Figure 7.13 (a) Control signal attached to reboiler; (b) attaching to controller.
Figure 7.14 (a) Selecting OP signal; (b) nal loop with signal connected.
Figure 7.15 (a) Initial controller faceplate; (b) tuning page tab.
The faceplate is located at some convenient spot in the window where we can keep an
eye on what is going on with this level loop. Remember that there will eventually be ve
controller faceplates.
Let us look in detail at the faceplate. As shown in Figure 7.17, there are six buttons at
the top. The rst button on the left is the Auto button, the second is Manual, and the third is
Cascade.
Figure 7.15 Continued.
Figure 7.16 Ranges page tab.
When the Auto button is pushed, the controller changes the OP signal automatically
according to the current values of the setpoint SP and the PV. The value of the setpoint
can be changed by double-clicking on the number in the box to the right of SP, typing
in the desired number, and hitting Enter on the keyboard.
When the Manual button is pushed, you can manually set the OP signal. This is done by
double-clicking on the number in the box to the right of OP, typing in the desired number,
and hitting Enter on the keyboard. When the Cascade button is pushed, the controller
receives its setpoint signal from some other control element. We will illustrate this later
in this chapter.
Now would be a good time to check out the pressure controller, which was automati-
cally set up when we started Aspen Dynamics. Figure 7.18a shows the faceplate (which
appears after double-clicking the icon on the owsheet) and the Congure page tab.
The default controller tuning constants are a gain of 20 and an integral time of 12 min.
These work pretty well in most column simulations. Note that the controller output is
not a % of scale signal sent to a valve. It is a heat removal rate in the condenser. As
a result, the controller is set up to be reverse-acting: when pressure goes up, the controller
output signal goes down. The reason for this action becomes clear when we look at the
Ranges page tab shown in Figure 7.18b. Note that the controller output ranges from a
minimum of 245,347,000 W to a maximum of 0 W. This conforms with the Aspen con-
vention that heat removal is negative.
In my experience the only modication that sometimes needs to be made to the pressure
controller is to change to a more convenient pressure transmitter range. For example, in
this column the operating pressure is 16.8 atm. We might change the pressure transmitter
range to 1419 atm from the very wide range used in the default setup. Of course, the gain
should be correspondingly reduced.
The second-level controller LC12 for the reux drum is installed and connected in the
same way. The PV signal comes from the level on stage 1. The OP goes to valve V12. A
direct-acting proportional-only controller is specied.
The nal basic controller that we need to set up is a ow controller on the feed. A PID
controller is placed on the owsheet. Its PV signal is the molar owrate of feedstream F1.
Its OP signal goes to valve V1. After opening the Tuning page tab and clicking the
Initialize Values button, we set the controller to be Reverse-acting, and use conventional
ow controller tuning (K
C
0.5 and integral time 0.3 min), as shown in Figure 7.19a.
The most common error in setting up the ow controller is to forget to specify Reverse
action. Since ow control is very fast and essentially algebraic, it seems to help the
Figure 7.17 Faceplate details.
numerical integrator to use some ltering in a ow controller. The Filtering page tab is
selected, the Enable ltering box is checked, and a small lter time constant (0.1 min)
is typed in (see Fig. 7.19b).
The owsheet now has four controller faceplates displayed, as shown in Figure 7.20.
We have one more controller to add, a temperature controller that holds the temperature
on a selected tray by adjusting the reboiler heat input.
It is important to clarify what is happening to the reux ow. The column icon does not
show the plumbing details of a reux drum, pump, and reux valve. As mentioned earlier,
the default condition in Aspen Dynamics is that the mass owrate of reux is constant,
unless otherwise adjusted. For example, if we wanted to control reux drum level with
reux owrate, the level controller OP signal would be connected to the column and
the Reux.FmR would be selected. The second common application would be if we
wanted to ratio the reux owrate to the feed owrate. We will illustrate these by
examples in this and later chapters.
A word of caution is appropriate at this point. During the initialization of controllers,
quirky things sometimes occur. There are some bugs in Aspen Dynamics that sometimes
set the OP signal at the wrong initial value (e.g., at 100 instead of 50%) or the PV at a value
Figure 7.18 (a) Pressure controller; (b) Ranges page tab.
not equal to the steady-state value. To circumvent these problems, switch the controller to
manual, type in the correct OP value, and run the simulation for a while. Then switch the
controller to automatic and run out to a steady state.
While we are on the subject of running, once the simulation runs out in time and con-
verges to a steady state, the le should be saved. It is convenient to save the le with time
set equal to zero so that it can be used to establish initial conditions for new runs. To do
this, make an Initialization run and then switch to Dynamic (but do not perform a run).
Click the Rewind button, which is the fth one from the right on the upper toolbar (see
Fig. 7.20). A window opens (Fig. 7.21) on which you can select the Initialization Run
as the Select rewind snapshot and then save the simulation le. Note that the Sim
Time is 0. You can then rewind to these conditions whenever you want to start
again at this steady state.
7.5 INSTALLING TEMPERATURE AND COMPOSITION CONTROLLERS
Installing temperature and composition controllers is somewhat more involved than instal-
ling level and ow controllers because of three issues:
1. We need to include additional dynamic elements in the loop. Temperature and
composition measurements have signicant inherent dynamic lags and deadtimes.
These should be incorporated in the control loop. This is necessary so that we use
realistic controller tuning constants and do not predict dynamic performance that
is better than reality.
2. The tuning of temperature and composition controllers is more involved than simply
using heuristics as is done for ow and level controllers. Some convenient and
Figure 7.19 (a) Flow controller setup; (b) ltering.
7.5 INSTALLING TEMPERATURE AND COMPOSITION CONTROLLERS 167
effective tuning procedure is required. One of the best methods is to run a relay
feedback test to nd the ultimate gain and ultimate frequency, and then use the con-
servative TyreusLuyben tuning settings.
3. The appropriate location for the temperature or composition sensor may not be
obvious. Some method for making a good selection must be used. As discussed
Figure 7.20 Controller faceplates.
Figure 7.21 Rewinding to Snapshot.
in Chapter 6, there are several ways to approach this problem. These include looking
at the shape of the temperature prole in the column, calculating steady-state gains,
and using SVD analysis.
7.5.1 Tray Temperature Control
Let us rst discuss using a temperature controller to maintain a tray temperature in the
column. Looking at the temperature prole in Aspen Plus, we see that stage 9 displays
a fairly steep slope. Its temperature is 337.36 K.
A controller is installed on the owsheet in the normal way, except that instead of using
a PID controller, we select a PIDIncr controller. The important difference between these
two is that the PIDIncr controller has a built-in relayfeedback test capability, which
make this dynamic test a breeze.
The PV is selected to be the temperature on stage 9. The OP is selected to be the
reboiler heat input QRebR. Figure 7.22a shows the controller faceplate and the Tuning
page tab after the Initialize Values button has been clicked. The normal controller
output is 27,166,000 W. The controller action should be set at Reverse because if the
tray temperature is going up, the reboiler heat input should be decreased. It is convenient
to change the range of the temperature transmitter from the default 273401 K to a more
convenient and narrower range of 320370 K, as shown on the Ranges page tab in
Figure 7.22b.
The program is run to make sure that everything works okay without a lag or a deadtime
in the loop. Now we back up and insert a deadtime element on the owsheet between the
column and the TC9 temperature controller. The reason for installing the controller
initially without the deadtime element is to avoid initialization problems that sometimes
crop up if you attempt to install the deadtime and the controller all in one shot.
Before we proceed, it might be wise to save some of the work. Since a fair amount of
time has been spent in setting up the faceplates and arranging them on the screen, we can
avoid having to do this again by clicking on Tools in the toolbar at the top of the screen and
selecting Capture screen layout. The window shown in Figure 7.23 opens, on which we
enter an appropriate name. When the program is restarted, the screen layout can be
reinstalled by going to the Exploring window, clicking Flowsheet, and double-clicking
on the icon in the lower Flowsheet Contents window with the name you provided.
Now let us install the deadtime element. The line from stage 9 temperature is selected.
Right-clicking, selecting Reconnect Destination, and placing the icon on the arrow point-
ing to the deadtime icon, we connect the input to the deadtime. A new control signal is
inserted between the deadtime and the controller. The deadtime icon then is selected.
Right-clicking, selecting Forms from the dropdown list, and selecting All Variables, we
open the window shown in Figure 7.24a. The DeadTime value is initially 0 min. Note
that the Input and Output values are set at a default number, not the actual 337.36 K
value. A deadtime of 1 min is entered, and performing an Initialization run lls in the
correct values, as shown in Figure 7.24b. The nal owsheet and controller faceplates
are shown in Figure 7.25.
Everything is ready for the relayfeedback test. Clicking the Tune button opens the
window shown in Figure 7.26a. We specify a Closed loop ATV as the Test method.
The default value of the Relay output amplitude is 5%, which is usually good. For a
very nonlinear column, the amplitude may have to be reduced.
Figure 7.22 (a) Initial installation of temperature controller; (b) Ranges page tab.
To start the test, click the Run button at the top of the screen and click the Start test
button on the Tune window. To be able to see the dynamic responses, click the Plot
button at the top of the controller faceplate. After several (46) cycles have occurred,
click the Finish test button. Figure 7.26b gives the results. The predicted ultimate gain
is 5.9, and the ultimate period is 4.5 min.
The timescale in Figure 7.26b is fairly small. To get good-looking plots, the plot time
interval must be reduced from the default value of 0.01 h. Aspen Dynamics calls this par-
ameter communication time, and it can be accessed by going to the top toolbar in the
Aspen Dynamics window and selecting Run and Run Options. A window opens on
which Communication Time can be set. A value of 0.0005 h was used to get the plots
shown in Figure 7.26b. This parameter does not affect the results of a dynamic simulation,
except for slowing it down somewhat. It only affects the appearance of plots.
Finally the Tuning parameters page tab is clicked, the TyreusLuyben Tuning rule
(Fig. 7.26c) is selected, and the Calculate button is pushed. The resulting controller set-
tings are gain K
C
1.84 and integral time t
I
9.9 min. The TyreusLuyben tuning
Figure 7.23 Saving screen layout.
Figure 7.24 (a) Deadtime All Variables table; (b) after initialization run.
formulas are
K
C

K
U
3:2
t
I
2:2P
U
These are loaded into the controller by clicking the Update controller button. Run the
simulation out in time for a while to see how well these settings work in terms of bringing
the column to steady state. In the next section we will subject the column to disturbances
and evaluate the performance of several control structures.
7.5.2 Reux to Feed Ratio
Before we illustrate the use of a composition controller, it might be instructive to show
how a reux to feed ratio structure is set up. In Chapter 5, steady-state calculations indi-
cated that a R/F ratio scheme should do a pretty good job of maintaining product purities
in the propane/isobutane system in the face of feed composition disturbances and, of
course, feed owrate changes.
The R/F structure is implemented by using a multiplier block. The input of this block is
the mass owrate of the feed, which is 52.513 kg/s. The properties of any stream can be
found by clicking on the steam name, right-clicking, selecting Forms, and then selecting
Results. Figure 7.27 gives the stream information for the feed F1.
Figure 7.25 Flowsheet with controller faceplate.
Figure 7.26 (a) Setting up the relayfeedback test; (b) relayfeedback test results; (c) calculated
controller settings.
Figure 7.27 Results table for stream F1.
The output of the multiplier block will be the mass owrate of the reux. To determine
the design value of this variable, we can click the column icon, right-click, select Forms,
and select Results. This opens the table shown in Figure 7.28a. The second line from the
bottom gives the reux mass owrate of 61.7877 kg/s. So the multiplier block should mul-
tiply the feed mass owrate by the number (61.7877/52.513) 1.1766.
A Multiplier control model is placed on the owsheet. A control signal is connected
from the feed F1 (mass owrate) to the multiplier (labeled R/F ratio in Figure 7.29).
This is Input1. A second control signal is connected from the multiplier output to the
column and connected to the blue arrow pointing to the line below the condenser. A list
Figure 7.28 (a) Results table for column; (b) choices; (c) entering ratio constant.
of alternatives opens, and the top one, Reux.FmR, is selected (see Fig. 7.28b). To set the
constant in the multiplier, the icon is clicked and right-clicked. Forms and All Variables
are selected, and the window shown in Figure 7.28c opens, on which we enter the number
1.1766 for Input2. In this example, one of the inputs is a constant. In other examples,
both of the inputs can vary with time.
The nal owsheet and controller faceplates are shown in Figure 7.29. We will
compare the performance of this control structure with some alternatives later in this
chapter. First we want to illustrate the use of a composition controller.
7.5.3 Composition Control
We want to compare tray temperature control with two types of composition control. In
both, the composition of the propane product is measured directly and controlled at
2 mol% isobutane. The rst type is direct composition control, in which a single PI
controller is used with reboiler heat input manipulated. The second type uses a cascade
composition-to-temperature control structure.
Composition measurement typically has larger deadtime and lags than does tempera-
ture control. We assume a 3-min deadtime in the composition measurement.
First, we add a PIDIncr controller to the owsheet and make the appropriate connec-
tions and do not use a deadtime, which will be added later. The controller should be set
to Reverse. The PV is the mole fraction of isobutane in the distillate stream. The OP is
reboiler heat input. A composition transmitter range of 00.05 mF isobutane is used, as
shown in Figure 7.30.
After the simulation is run, a 3-min deadtime is inserted. Initialization and Dynamic
runs are made to converge to steady-state conditions. Then a relay-feedback test is run.
Results are shown in Figure 7.31. Note that the timescale on the plot is very different
Figure 7.29 Flowsheet with R/F structure.
Figure 7.30 Ranges page tab for cascade control.
Figure 7.31 Relayfeedback test.
from that for the temperature controller. The ultimate gain is 0.58, and the ultimate period
is 32.4 min. The TyreusLuyben settings are calculated and inserted in the composition
controller. The nal owsheet is given in Figure 7.32.
7.5.4 Composition/Temperature Cascade Control
Temperature control has the advantage of being fast, but it may not hold the product purity
constant. Composition control is slow, but it will drive product purity to the desired value.
The nal control structure studied in this chapter is a cascade combination of composition
and temperature control that achieves both fast control and the maintenance of product
purity.
The tray temperature controller is the secondary controller. It is set up in exactly the
same way as we did in the previous section. It looks at tray temperature and manipulates
reboiler heat input. However, its setpoint is not xed. The setpoint signal is the output
signal of the composition controller, which is the primary controller.
The tuning of the secondary temperature controller remains unchanged. The primary
composition controller must be retuned since its ouput signal is now a temperature
setpoint. With the temperature controller set on automatic, the relayfeedback test is
run on the composition controller. Figure 7.33 shows the relay feedback test results.
The ultimate gain and ultimate period are 0.98 and 15.9 min, respectively, compared to
the direct composition results of 0.58 and 32.4 min. We can see immediately that a
higher gain and smaller integral time result, which indicates tighter control with the
cascade control structure. Figure 7.34 shows the cascade control structure and controller
faceplates. Note that the TC7 temperature controller is on cascade (meaning that its set-
point signal is the output signal of the composition controller).
Figure 7.32 Final owsheet with cascade control.
7.6 PERFORMANCE EVALUATION
We want to see how well the three alternative control structures developed above perform
in the face of disturbances, specically, how close to the desired values of temperature and
composition these variables are maintained, both at steady state and dynamically. A dis-
turbance is made and the transient responses are plotted.
Figure 7.33 Cascade control relayfeedback test.
Figure 7.34 Cascade control.
7.6 PERFORMANCE EVALUATION 179
7.6.1 Installing a Plot
To see what is going on, the rst thing to do is to set up a plot or a stripchart. This will show
howthe variables of interest change dynamically with time. To open a plot, go to Tools on the
top toolbar and click New Plot. Aplot windowopens, as shown on the left in Figure 7.35a. A
number of variables will be plotted: owrates of the feed, distillate, bottoms, and reux; stage
7 temperature; reboiler heat input; and distillate impurity (mf isobutane) and bottoms impur-
ity (mf propane). We drop and drag each one of these variables onto the plot window. The
table on the right of Figure 7.35a is opened by clicking the icon for the F1 stream, right-click-
ing, and selecting Forms and then Results. Place the cursor just to the right of F (total molar
ow) in this table and click once. Then click and hold the left mouse button, drag the cursor to
the plot window, and release. This procedure is repeated for each variable. The plot with all
variables added is shown in Figure 7.35b. Now we need to dene the scales of each variable,
simplify the labels, and make other beautication changes to the plot. This is done by right-
clicking on the plot and selecting Properties. The windowshown in Figure 7.36 opens, which
contains a number of page tabs. On the Axis Map page tab, click the box One for Each to get
different scales, which are then set on the Axis page tab. If you want xed scales, uncheck the
box to the left of Reset axis range to data for each of the variables and dene minimum and
maximum values for each.
Figure 7.37 illustrates the type of plot that results from making a dynamic run. At time
equal 0.1 h, the setpoint of the feedow controller is changed from 1 to 1.2 kmol/s. To
Figure 7.35 (a) New plot and F1 results table; (b) plot with variables.
Figure 7.36 Properties of plot.
Figure 7.37 Plot for 20% feedow; cascade.
pause the simulation at some point in time, go to the top toolbar and click Run and then
Pause At, which opens the window shown in Figure 7.38.
Although these Aspen Dynamics plots are useful to see what is going on during each
run, they have a xed structure (multiple variables per plot) and do not permit comparisons
of results from different runs. Fortunately, this is easily handled by storing the data in a le
and then producing your own plots using your favorite plotting software. To store the data,
right-click the plot and select Show as History. Figure 7.39 shows the resulting table,
which can be copied and pasted into a le.
Runs are made for all three control structures with two disturbances. Feed owrate is
changed by changing the feedow controller setpoint. Feed composition is changed by
clicking on the F1 stream icon, right-clicking, and selecting Forms and then Manipulate.
Figure 7.38 Pause window.
Figure 7.39 Show as History table.
The window shown in Figure 7.40 opens, on which the mole fractions of propane and
isobutane are changed from 0.40/0.60 to 0.50/0.50.
The data are loaded into Matlab for plotting. Figure 7.41 compares the responses for a
20% increase in feedow. The solid lines indicate when only a tray temperature is con-
trolled. The dashed lines show when the composition/temperature cascade control struc-
ture is used. The direct composition control structure results are not shown because the
responses are very poor. This is as we would expect because of the very low controller
gain and the very large integral time.
The cascade control structure gives tighter control with smaller peak transient disturb-
ances and a shorter settling time. Both control structures return the product purities to their
desired values because of the R/F ratio used in both.
Figure 7.42 gives results for feed composition disturbance. At time equal 0.1 h, the feed
composition is changed from 40 mol% propane to 50 mol% propane. The temperature
control structure holds the temperature on stage 9 constant. Both distillate and bottoms
Figure 7.40 Manipulate window.
0 0.5 1 1.5 2
335
335.5
336
336.5
337
337.5
338
T
r
a
y

9

(
K
)
0 0.5 1 1.5 2
26
28
30
32
34
36
Q
R
(
M
W
)
0 0.5 1 1.5 2
1.6
1.7
1.8
1.9
2
2.1
x
D
(
%

i
C
4
)
Time (h)
0 0.5 1 1.5 2
0.95
1
1.05
1.1
1.15
1.2
x
B
(
%

C
3
)
Time (h)
TC
CC/TC
CC/TC
CC/TC
CC/TC
TC
TC
Figure 7.41 TC and cascade; feedow disturbance.
impurities change signicantly. The distillate impurity climbs from the desired 2 mol%
isobutane level to over 3 mol%. The bottoms stream becomes overly pure. Using
the CC/TC cascade control structure, the distillate purity is maintained. The setpoint of
the temperature controller is decreased by the composition controller to drive the distillate
impurity to 2 mol% isobutane.
These results illustrate that single-end temperature control provides stable base-level
regulatory control of the column. However, product purities change somewhat when
feed composition changes.
7.7 COMPARISON WITH ECONOMIC OPTIMUM DESIGN
Now that all the basic dynamic simulation tools have been developed and their application
illustrated, the last issue to be covered in this chapter is the effect of column design on
control performance. In Chapter 4 the economic optimum design for this system was
shown to be a 44-stage column operating with a reux ratio of 3. So far in this chapter
a 32-stage column with a 3.46 reux has been simulated and controllers developed.
This section provides a comparison of these two alternative designs. Is the economic
optimum design also the best design from the standpoint of dynamic controllability?
A dynamic simulation of the 44-stage column is developed. The smaller column diam-
eter (5.22 vs. 5.91 m) and lower reboiler heat input (24.8 vs. 27.2 MW) gives a slightly
smaller reux drum (D 4 m) and base (D 5 m). Stage 9 is selected for temperature
control (336 K at steady state).
A relayfeedback test gives an ultimate gain of 5.31 and an ultimate period of 4.8 min.
The corresponding values for the 32-stage column are 5.9 and 4.5 min. These results
0 0.5 1 1.5 2 2.5 3
336
336.5
337
337.5
T
r
a
y

9

(
K
)
0 0.5 1 1.5 2 2.5 3
26.8
27
27.2
27.4
27.6
27.8
28
Q
R
(
M
W
)
0 0.5 1 1.5 2 2.5 3
1.8
2
2.2
2.4
2.6
2.8
3
3.2
x
D
(
%

i
C
4
)
Time (h)
0 0.5 1 1.5 2 2.5 3
0.6
0.8
1
1.2
x
B
(
%

C
3
)
Time (h)
TC
CC/TC
CC/TC
CC/TC
CC/TC
TC
TC
TC
Figure 7.42 TC and cascade; feed composition 4050% C3.
0 0.5 1 1.5 2
333
334
335
336
337
338
T
r
a
y

9

(
K
)
0 0.5 1 1.5 2
24
26
28
30
32
34
Q
R
(
M
W
)
0 0.5 1 1.5 2
1.6
1.7
1.8
1.9
2
2.1
x
D
(
%

i
C
4
)
Time (h)
0 0.5 1 1.5 2
0.95
1
1.05
1.1
1.15
1.2
x
B
(
%

C
3
)
Time (h)
44
32
44
44
44
32
32
32
(a)
0 0.5 1 1.5 2
336
337
338
339
340
T
r
a
y

9

(
K
)
0 0.5 1 1.5 2
18
20
22
24
26
28
Q
R
(
M
W
)
0 0.5 1 1.5 2
2
2.2
2.4
2.6
x
D
(
%

i
C
4
)
Time (h)
0 0.5 1 1.5 2
0.95
1
1.05
1.1
1.15
x
B
(
%

C
3
)
Time (h)
44
32
44
44
44
32
32
32
(b)
Figure 7.43 Columns with 32 and 44 stages: (a) 20% feedow; (b) 220% feedow.
7.7 COMPARISON WITH ECONOMIC OPTIMUM DESIGN 185
indicate that somewhat tighter control can be achieved in the 32-stage column because of
the larger controller gain and smaller integral time.
These predictions are not borne out by dynamic simulations. Figure 7.43a gives results
for the two columns with a 20% step increase in feedow. For this disturbance, the biggest
difference between the two columns is in the bottoms impurity. The 44-stage column dis-
plays a much smaller transient peak departure from the desired value of 1 mol% propane.
Although the response of the 32-stage column is faster, the dynamic deviation is larger.
The explanation for this difference is that there are more trays between the feedpoint
and the base (feed stage is 19 and base is 44, giving 25 trays) in the 44-stage column.
In the 32-stage column the feed stage is 14 and the base is 32, giving only 18 trays. There-
fore the disturbance takes longer to reach the base and the temperature controller has more
time to take corrective action.
We could try to advance the general conclusion that having more trays in the column
helps the dynamic response. This certainly is the conventional wisdom in distillation
control.
However, the response of the system to other disturbances brings this conclusion into
question. Figure 7.43b gives results for the two columns with a 20% step decrease in feed-
ow. Now the deviation in bottoms composition is larger in the 44-stage column and the
response is quite slow.
It should be remembered that a reux to feed ratio is adjusting the reux ow to
the top of the column as soon as the feed owrate changes. The change in reux affects
the control tray quickly because it is only eight stages down in both columns. For an
increase in feed owrate, the increase in reux starts to reduce the temperature on the
control tray, and the temperature controller increases the heat input, which affects
the bottoms composition quickly. So the bottoms composition is being affected by three
0 0.5 1 1.5 2 2.5 3
335.5
336
336.5
337
337.5
T
r
a
y

9

(
K
)
0 0.5 1 1.5 2 2.5 3
23
24
25
26
27
28
Q
R
(
M
W
)
0 0.5 1 1.5 2 2.5 3
2
2.5
3
x
D
(
%

i
C
4
)
Time (h)
0 0.5 1 1.5 2 2.5 3
0.6
0.8
1
1.2
x
B
(
%

C
3
)
Time (h)
44
32
44
44
44
32
32
32
Figure 7.44 Columns with 32 and 44 stages; feed composition 4050% C3.
inputs: the feed owrate, the reux owrate, and the reboiler heat input. These inputs all
have different timescales in terms of the time when they affect the bottoms composition.
The net effect is difcult to predict. This demonstrates the usefulness of a rigorous
dynamic simulation.
Figure 7.44 gives results for a feed composition change. The difference between the
two columns is quite small. There are no changes in reux or feed owrates. So the vari-
ables affecting bottoms composition are the feed composition and the vapor boilup
changes that result from the feed composition affecting the control tray temperature.
Note that the energy consumption in the 32-stage column is higher than that in the
44-stage column.
These results demonstrate that distillation column dynamics are not as simple as one
might expect, even in this ideal VLE system. The dynamic interplay between the
effects of disturbances, measurements, and manipulated variables can lead to some
unexpected and counterintuitive results.
7.8 CONCLUSION
A very full bag of distillation dynamic simulation techniques has been developed and
demonstrated in this chapter. The example considered is a simple binary ideal VLE
column. As the remaining chapters in this book demonstrate, these techniques can be
readily extended to much more complex owsheets and phase equilibria.
7.8 CONCLUSION 187

CH 07

Uploaded by

Copyright:

Available Formats

CH 07

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CH 07

Uploaded by

Copyright:

Available Formats

CHAPTER 7

CONVERTING FROM STEADY STATE

You might also like