2016
elements of H: for small s
412B
hl2 =
(1 + 411c1)(1+ q22c2)
421 4
h2l = For large s, the assumption (qll(s)q22(s)( >>
(1+ 411c1)(1 + 422c2)
Iql2(s)q2l(s)Jused to derive H is still reasonable, be
Due to the integral action in the controllers, the in cause this means that the diagonal elements of G is
teraction is small at low frequencies. To estimate the dominating at high frequencies. Hence, the transfer
maximum of the interaction, we observe that function H describes the relation between Y, and Y
also for large s. This leads to that
hi2 = Q I ~ & & Sh21=
~ , ~ ~ I E I S I S ~(1)
where SI = ( l + q l l q )  l and S2 = ( l + q 2 2 c ~ )  I are the
sensitivity for loop one and two, respectively, if the in
teraction is neglected. Upper bounds of the interaction for large s, where d1 is the pole excess of q12(s) and
terms are thus given by d2 is the pole excess of 421(s).. Note that dl and d2
are uniquely defined by G, and do not depend on the
Ih12(iw)l 5 lq12~2l~slMs2 controllers. It follows from (3) and (4) that Ihl~(iw)I
Ih2l(iW)I I 1q21%IMs1Ms2 and Ihzl(iw)) will have maxima. The location of the
where Msl and Ms2 are the maximum sensitivity for maximum of Ihlz(2w)l can, by using (3) and (4), be
the individual loops. For low frequencies, we have approximated by the solution of the equation
and
E1 x kll/s, E2 x k12/~ The peaks of the diagonal terms Ihlz(iw)l and (h21(iw)I
For low frequencies, we thus find that 41222 and 421E1 are thus approximately located at
are constant. It is therefore natural to introduce the
interaction indices
2017
I 0.5
0 10 20 M 40 50 60 70 80 0 10 20 30 40 SO 60 70 80
t t
Yz YZ
4 5
0 10 20 30 40 50 60 IO
I
60
t
Figure 1: Simulation of the design method applied to Figure 2: Simulation of controllers without setpoint
Rosenbrock's system. The figure shows the weighting. The simulation is identical to Fig
response of the outputs to steps in the com ure 1 but the controllers have bl = bp = 1.
mand signals. The PI controllers have setpoint
weighting bl = bz = 0.
a zero s = 1 in the right halfplane. If we introduce
static decoupling, the compensated transfer function
optimization technique which minimizes integral gain becomes
subject to a robustness constraint. This is discussed
for PI controllers in [3] and for PID controllers in [lo].
Such a method will give controllers which are optimized
with respect to rejection of load disturbances. The de
sign methods will give integral gains kyk > 0, which
US)
= i o "1
3(1  S)
(s + 1)(: +3) (s +sl)(s
1l
+
+ 3)
could be used if there were no interactions. The ratios The interaction is given by 1 ~ 1 2= 4/3 and ~ 2 1= 0.
Ak = k I k / k y k are thus measures of performance losses Since l ~ z l= 0, interaction gives no performance limi
due to the interaction. If A k 2 1 there are no perfor tations for the second loop. There are however limi
mance losses and the limitations on performance are tations because of the right halfplane zero at s = 1.
essentially given by the loop dynamics. If 0 5 X k < 1 Designing a PI controller that maximizes integral gain
it is necessary to detune the controllers. In this case we subject to the constraints that the maximum sensitivity
need methods to design PID controllers with specifica Msl and the maximum complementary sensitivity Mpl
tions on the closedloop bandwidth. There are many are less than fi,gives k p l = 0.245 and k I l = 0.248,
methods which can be used for this purpose. Two sim see [3]. Since 1 ~ 1 2= 4/3 there are constraints on the de
ple techniques for design of PID controllers are the sign of the first loop because of the coupling. Requiring
direct poleplacement design based on reducedorder that the coupling ICI is less than 0.1 and the maximum
models and the dominant pole design, see [2] and [ll]. sensitivity M,z is less than fi9 we find that the in
tegral gain of the second loop IC12 must be less than
~ l / ( n l 2 M , 1 M , z )= 0.0750. To design a PI controller,
we use direct pole placement [2] based on the model
3 ExampIes +
qz2(s) = (s 1)l. This gives kp2 = 2f,uo  1 and
IC12 = WO". Requiring that the integral gain is equal to
The control design method is in this section illustrated 0.0750, we find that WO = 0.274. With C = 1 the con
on two examples from the literature. troller gains become kp2 = 0.452 and k ~ = z 0.075.
Figure 1 shows simulations of setpoint responses for
Rosenbrock's System the closedloop system. The plots show the proposed
The process design with setpoint weighting (bl = b2 = 0). A unit
step in the set point of the first controller is applied at
time t = 0 and a step in the set point of the second con
G(s)= Isll I troller is then applied at time t = 40. Figure 2 shows
\s+l s+lj the step responses for a controller without setpoint
weighting. The figure clearly indicates the advantage
was originally proposed by Rosenbrock [13]. It is an of setpoint weighting for multivariable systems. The
example of a system that looks very easy to control, reason why there is such a large difference is that the
but which has fundamental limitations because it has control signal is much smoother with setpoint weight
2018

lo"
.e&'
10 4 100
w
\
L,Kl

102
fF
lop 10"
W
1o=
A series
~ 2 =
y:i 1  17.39
0.5138s giy:2.31s
10'
W
10' 10 10"
w WO2 < Jm= do = 0.022
Figure 3: The frequency response of the closedloop sys We have wolL = 0.36 and wozL = 0.066. The band
tem with setpoint weighting (solid) and with widths are thus well below the limitations w L < 0.7
Out (dashed). Note without setpoint imposed by the the largest time delay of the system
weighting the interaction Ihlz(iw)l is larger and ( L = 3), see [I].
extends to higher frequencies.
With these frequencies and the time delays present in
ing. the system, it is quite reasonable to approximate the
time delays with first order lags. The model reduction
The effect of setpoint weighting
.~
is illustrated also in should thus work reasonably well in this case. Choosing
Figure 3, which shows the frequency response of the (1 = (2 = C = 0.707 we find that the proportional gains
closedloop system with (solid) and without (dashed) are
setpoint weighting. The interaction increases consid
erably when no setpoint weighting is applied. This kpi = 2Cwol  1 = 1.1338
agrees with the conclusion in previous section, com lcp2 = ~ ( W O Z 1 =z 0.4699
pare the interaction indices with and without setpoint
weighting in Equations (2) and (6). In Figure 4 we show the step responses from set point
t o process output for the system. Analyzing the figure
The maximum of the interaction 1&2(iw)( in Figure 3 we find that the response in y1 caused by a step in the
is equal to 0.3. This should be compared t o the inter set point yr2 is quite small. It thus seems possible t o
action index 6 1 = 0.2. The maximum is attained at increase the bandwidth in the second loop. After some
w = 0.52, while the estimate in (5) predicts 0.41. experimentation we found that a reasonable value is
w02 = 0.1, which gives that the controller parameters
It is interesting to note that the bandwidth of the for the second loop become k p 2 = 0.7 and klz = 0.123.
closedloop system (as measured by the maximal sin The time responses are shown in Figure 5. A compari
gular value) is equal to wb = 0.49 rad/s, which is fairly son with Figure 4 shows that the response speed of the
close to the limitation imposed by the right halfplane second loop has been increased considerably without
zero at +l. increasing the interaction too much. This also shows
that the method lends itself very well to tuning. The
WoodBerry's Binary Distillation Column step responses in Figure 5 compare favorably with de
The WoodBerry binary distillation column plant [14] signs obtained by other methods.
is a multivariable system that has been studied exten
sively. The process has the transfer function
/ 12.8eS 18.9e3"\ 4 Conclusions
2019
Y1
I 1
Y2 Y2
0.5l I 0.d J
50 100 150 200 250 3W 50 100 150 200 250 300
t t
Figure 4: Simulation of the design method applied to Figure 5: Simulation of the design method applied to the
the WoodBerry distillation column. The figure WoodBerry distillation column. The band
shows the response of the outputs to steps in width of the second loop has been increased
the command signals. to WO2 = 0.1.
assumption that the interaction is not too severe. Pos [7] W. L. Luyben. Process modeling, simulation,
sible detuning of the PID controllers was quantified via and control for chemical engineers. McGrawHill, New
the interaction indices I E ~and 62. An important remark York, 1990.
made in the paper was that setpoint weighting is neces
[8] J. M. Maciejowski. Predictive Control with Con
sary for PID controllers in multivariable systems. The straints. Pearson Education, 2000.
interaction between the control loops can be reduced
considerably if setpoint weighting bl = bz = 0 is used. [9] A. Niederlinski. A heuristic approach to the de
sign of h e a r multivariable interacting control systems.
Other methods for doing detuning for multivariable Automatica, 7:691701, 1971.
PID control includes Niederlinski’s heuristic design [9]. [lo] H. Panagopoulos. PIDControl. Design, Ezten
Note that the interaction indices tc1 and IEZ depend on sion, Application. PhD thesis, Dept. of Automatic Con
the controller, which is for instance not the case for the trol, Lund University, Sweden, 2000.
widely used interaction measure RGA by Bristol [4].
[ll] P. Persson and K. J. Astrom. Dominant pole
design  a unified view of PID controller tuning. In
Preprints 4th IFAC Symposium on Adaptive Systems in
References Control and Signal Processing, pages 127132, Greno
[l] K. J. Astrom. Limitations on control system per ble, France 1992.
formance. European Journal on Control, 6:220, 2000.
[12] S. J. &in and T. A. Badgwell. An overview of
[2] K. J. Wstrom and T. Hagglund. PID Controllers: industrial model predictive control technology. In J. C.
Theory, Design, and Tuning. Instrument Society of Kantor, editor, Conference on Chemical Process Con
America, Research Triangle Park, NC, 1995. trol CPC V, pages 232256, London, 1996. CACHE,
[3] K. J. bstrom, H. Panagopoulos, and Elsevier.
T. Hagglund. Design of PI controllers based on [13] H. H. Rosenbrock. StateSpace and Multivariable
nonconvex optimization. Automatica, 34(5):585601, Theory. Nelson, London, U.K., 1970.
1998.
[14] R. K. Wood and M. W. Berry. Terminal compo
[4] E. Bristol. On a new measure of interaction for sition control of a binary distillation column. Chem.
multivariable process control. IEEE l h n s . Autom. Engng. Sci., 28:17071717, 1973.
Control, 11:133, 1966.
[5] E. F. Camacho and C. Bordons. Model Predictive
Control. Springer, London, 1999.
[6] Y. Lee, S. Park, and J. H. Lee. On interfacing
model predictive controllers with lowlevel loops. In
Preprints. 14th World Congress of IFAC, volume N,
pages 313318, Beijing, China, 1999.
2020