Abstract: This paper presents the simulation of a simple First Order plus Delay Time (FOPDT) process model using
advanced control algorithms. Specifically, these advanced algorithms are the IMC-based PID controller, the Model
Predictive Controller (MPC) and the Proportional-Integral-Plus Controller (PIP) and their performance is compared
with the conventional Proportional – Integral – Derivative (PID) algorithm. The simulations took place using the
Matlab/Simulink™ software.
Key-Words: FOPDT, advanced control algorithms, IMC-based PID, MPC, PIP, PID, Matlab/Simulink™
Controller K P K c TI TD λθ
IMC-based τ θ
PI without τ >1.7
λ 2
Padé
IMC-based 2τ θ θ
τ >1.7
PI 2λ 2
IMC-based 2τ θ θ τθ
τ >0.8
PID 2λ θ 2 2τ θ
Table 2: IMC-based PID tuning parameters of a
FOPDT process
where
1.4 PIP Controller Q is a n m n m weighting matrix and
PIP controller comprises a part of the True Digital
R is a scalar input u (i ) weight
Control – TDC control method and can be considered
as a logical extension to the conventional PI/PID
controller but with inherent model predictive control It is worth noting that, because of the special
action. The power of the PIP design derives from its structure of the state vector x(k), the weighting matrix
exploitation of a specialized Non-Minimal State Space Q is defined by its diagonal elements, which are
(NMSS) representation of a linear and discrete system directly associated with the measured variables and
referred as NMSS/PIP formulation [13] [14]. integral-of error state vector. For example the diagonal
The fact that the PIP is considered as a logical matrix can be defined in the following default form.
extension of the conventional PI/PID controlled can be
appeared better when the process‟s transfer function is Q diag q1 qn qn 1 qn m1 qn m (12)
second order of higher or includes transport lag greater qy 1 n
qu 1 m qe
than one sampling interval. Then PIP controller
The SVF gains are obtained by the steady-state
includes also a dynamic feedback and input
solution of the well-known discrete time matrix
compensation introduced “automatically” by the
Riccati Equation [17], given the NMSS system
specialized NMSS formulation of the control problem
description (F and q vectors) and the weighting
[15] that in general, has a numerous advantages
matrices (Q and R).
k f 0 f1 f n 1 g1 g m 1 k1
against other advanced control structures [16].
(13)
Any linear discrete time and deterministic SISO In a conventional feedback structure, the SVF
ARIMAX model can be represented by the following controller can be implemented as shown in Figure 4,
specialized NMSS equations. where it becomes clear how the PIP can be considered
x(k ) Fx(k 1) qu(k 1) dyd (k ) (6) as a logical extension of the conventional PI/PID
algorithm, enhanced by a higher-order forward path
y(k ) hx(k ) (7)
and input/output feedback compensators
where the vectors F , q , d and h comprise the
parameters of the above equations [14]
G [ z 1 , m 1] and F [ z 1 , n 1] respectively disturbances or not, the „default‟ LQ weight matrices
[15]. for the PIP controller are;
Qdiag 1 0.25 0.25 0.25 1 , R 0.25 ,
F ( z 1) f 0 f1z 1 f n 1z (n 1) (14)
(absence of measured disturbances) and
G( z 1 ) 1 g1z 1 g m 1z (m 1) (15) Qdiag 1 0.25 0.25 0.25 1 0 0 , R 0.25 ,
(presence of measured disturbances).
3. Problem Solution
With no disturbances and input constraints, the
output response (Figure 6) for the advanced control
algorithms yields satisfactory step behavior with good
set point tracking and smooth steady state approach.
However, the response of the conventional PID seems
Figure 4: PIP feedback block diagram to be rather disappointing, as it yields a large
overshoot. Figure 7 demonstrates their control action
response. Mainly concerning MPC and PID
2. Problem Formulation algorithms, the initial sharp increase of their control
In order to asses the practical utility of the above action signal may not be acceptable during a practical
described advanced control algorithms, a series of realization of the controller in an actual industrial
implementation simulations have been conducted on a plant.
simple FOPDT process. For comparison purposes, a Figure 8 shows the output response after the
conventional PID controller is also designed using the introduction of input constraints defined by (17).
Ziegler-Nichols method. According to the results, both PIP and IMC-based PID
The FOPDT process model is described by (16) controllers were unaffected by the input constraints as
and initially is assumed absence of plant model their constrained control action response has been
mismatch, inputs constraints or measured disturbances. within the constrained limits. Although the response of
The model selection is based on the fact that a FOPDT the conventional PID controller retained its large
model represents any typical SISO chemical process. overshoot, the introduction of input constraints has
The simulation took place using the Matlab/Simulink™ optimized its smoothness. Finally MPC maintained its
software and the results are discussed in terms of Type satisfactory performance, although the fact that its
1 servomechanism performance. manipulated variable has been constrained the most
1 0.3s (Figure 9).
G( s) e Figure 10 demonstrates the output responses of the
s 1 (16)
process during the introduction of measured
The next simulation scenario includes constraints in disturbances defined by (18). According to the results,
the input manipulated variables. MPC controller yields the most optimal response while
2 u(t ) 2 (17) PIP controller sustains its performance. On the
In the final simulation scenario a simple contrary IMC-based PID as well as the conventional
disturbance model described by (18) is also PID yield a rather large overshoot.
implemented, in order to study the capability of each Table 3 shows an approximate numerical
controller in disturbance rejection. evaluation of the control algorithms for each scenario.
0.8 0.1s (18) The evaluation parameters are the Overshoot (O), Rise
Gd ( s) e
s 1 Time (RT), Settling Time (ST), Integral Square Error
The critical characteristics for the estimation of PID (ISE), Robust stability (RS) and Robust Performance
parameters (See Table 1) are Kcr=5.64 and (RP).
Pcr=1.083. The IMC-based PID parameters are
estimated according to Table 2 selecting 0.5 and
n 1. The calculation of MPC gain matrix includes
the following parameters; input weight Γlu 1 , output
weight Γly 0 , control horizon m 10 and infinity
prediction horizon. Whether the absence of measured
Controller %O RT ST ISE
Scenario 1
PID 49.80 0.5300 1.9300 0.49
IMC-based PID 1.76 1.1800 1.3800 0.51
MPC 0.00 0.0021 0.0021 ???
PIP 0.00 1.2500 1.4500 0.65
Scenario 2
PID 50.00 0.9700 2.6600 0.67
IMC-based PID 2.00 1.2400 1.4400 0.52
MPC 0.00 0.8500 0.9500 ???
PIP 0.00 1.2500 1.4500 0.65
Scenario 3
PID 62.95 0.5300 1.9300 0.48
IMC-based PID 16.23 0.7800 3.3800 0.40 Figure 7: Output Step Response with Input Constraints
MPC 0.00 0.0021 0.0021 ???
PIP 7.38 0.9500 1.9500 0.50
Table 3. Numerical Evaluation of Control Algorithms
[3] Astrom K., J., & Haugglund T. PID [15] Taylor C., J., Leigh, P., Price, L., Young, P., C.,
Controllers: Theory, Design and Tuning. Vranken, E., Berckmans, D., Proportional-
Instrument Society of America, 1995 Integral-Plus (PIP) Control of Ventilation
Buildings. Pergamon Control Engineering
[4] Coughanour, D. Process Systems Analysis & Practice, 2003.
Control NY, McGraw-Hill c1991 Chemical
Engineering Series, 1991 [16] Taylor, C., J., Chotai, A., Young, P., C., State
space control system design based on non-
[5] Rivera, E., D., Internal Model Control: A minimal state-variable feedback: further
Comprehensive View. Arizona State University, generalization and unification results.
Arizona, 1999 International Journal of Control, Vol. 73, No
14, 1329-1345, 2000.
[6] Bequette, B. W., Process Control: Modelling,
Design and Simulation. Prentice Hall, Upper [17] Astrom, K., J., & Wittenmark, B., Computer
Saddle River, NJ, 2003 controlled systems: Theory and Design.
Prentice-Hall Information and System Sciences
Series, 1984