Performance assessment of control loops

International Journal of Science and Research (IJSR), India Online ISSN: 2319-7064
Volume 1 Issue 2, November 2012
www.ijsr.net
Performance Assessment of Control Loops
Partha Saha1
, Chinmoy Basak Mukta2
, Dr M.A.A. Shoukat Choudhury3
1
Bangladesh University of Engineering and Technology, Department of Chemical Engineering,
Dhaka-1000, Bangladesh
sarothi16che@gmail.com
2
Bangladesh University of Engineering and Technology, Department of Chemical Engineering,
cmbasak@gmail.com
3
Bangladesh University of Engineering and Technology, Department of Chemical Engineering, Process Control,
shoukat@che.buet.ac.bd
Abstract: The main concern of this thesis is to assess control loop performance using minimum variance control by FCOR algorithm in
a convenient way by using only routine operating data and comparing the outcomes with an existing algorithm (named PINDEX). Only
by determining a performance factor of the feed-back control loop the controller performance could be assessed, even from thousands of
control loops in process industries. Minimum variance is the benchmark which indicates attainable minimum variance against the
actual variance of the control error. A performance index has been defined as the ratio of minimum variance to actual variance. Where,
a performance index value close to 1 indicates good control a value close to 0 indicates poor control. The performance index is always
bounded between 0 to1. The performance index calculation algorithm was implemented using MATLAB. A dynamic simulation study
was performed using Aspen HYSIS for generating process data and also by SIMULINK. Various cases such as noise corruption, stiction
and oscillation were introduced and the aftermath was studied and analyzed. Industrial data from an Ammonia plant (SAFCO) was also
analyzed. All the results are compared with previously developed algorithm. This thesis will help to outline a very simple but effective
way to detect performance of the controller in process systems.
Keywords: FCOR, SISO, Performance Index, Feed-back Controlling.
1. Introduction
In any chemical complex the foremost endeavor for any
chemical engineer is to-make money save money. In
process controlling it is very indispensable to get some
quick idea that all the controller in control loops
working satisfactorily or not. If this checking upon
collection of process data could be done within a very
short time then the frequency of online checking could
be enhanced. So it is very important to take decision
about replacing any existing controller or keeping it in
accordance with its performance factor. As a relevance
to this pragmatic demand, the main objective is to assess
control loops performance by applying minimum
variance control law, using closed loop process data.
According to minimum variance controlling minimum
variance is the benchmark which indicates attainable
minimum variance against the actual variance of the
control error. A performance index has been defined as
the ratio of minimum variance to actual variance of the
control loop. Thus performance index value close to 1
indicates good control while value close to 0 indicates
poor control. The performance index value is always
bounded between 0 to1. For performance index
evaluation, algorithm implemented using MATLAB.
2. About Methodology
A controller performance assessment technique has been
developed using the routine operating data for univariate
control loops, assuming that control objective is to
reduce process variance; this conventional approach is
termed as Minimum Variance Control. It is used
naturally as the benchmark standard against which
current control loop performance is assessed. It has been
acknowledged that a system with time delay d, a portion
of output variance is feedback control invariant and can
be estimated from routine operating data. This portion of
output variance equals the variance achieved under
minimum variance control; thus the method for the
estimation of the minimum variance from routine
operating data is established. Observed minimum
variance for any system means theoretically achievable
absolute lower bound of output variance to assess
control loop performance. Using minimum variance
control as the benchmark does not mean that one has to
implement such a controller on the actual process. This
benchmark control may or may not be achievable in
practice depending on several physical constraints.
However, as a benchmark, it provides useful
information such as how well the current controller and
how much potential there is to improve controller
performance further. If the controller indicates a good
performance measure relative to minimum variance
control, further tuning or re-designing of the control
algorithm is neither necessary nor helpful. In this case, if
further reduction of process variation is desired,
implementation of other strategies such as feed forward
control or re-engineering of the process itself may be
necessary. On the other hand, if the controller indicates
a poor performance measure, further analysis, such as
model analysis, robustness analysis, constraint analysis
etc may be necessary.
8

www.ijsr.net
3. Experimental Work done
3.1 Performance Assessment of MATLAB (SIMULINK)
model (without stiction block)
Figure 1. MATLAB Simulink Process block for process
data generation.
According to the simulated process Figure 1 above process
data is generated. As noise was added into the model process
variable shows those uncorrelated noise of the model. Later
on plot of process variable with data points with was
generated to demonstrate the process behavior. The output of
the controller was also observed. Then process data was
taken for consideration by both the implemented algorithm
and the previously developed one.
Figure 2. Plot of process variable vs. time
Figure 3. Plot of controller output vs time
Figure 2 shows generated data at different data points.
Figure 3 shows controller output at different data points.
Controller is basically used to keep the process variable
within certain range. To do this controller variable constantly
changes with time.
3.1.1 Results
Results found form Pindex and FCOR algorithm are shown
below-
Pindex=0.1494
FCOR=0.1503
Performance index is found from study according to
minimum variance control law. It is evident from the result
that the control loop is performing poorly (as the index
values are close to 0). Both Pindex and FCOR algorithm
show similar results. It assures that developed algorithm
works nicely to evaluate the performance index of this
simulated process without valve stiction of the controller.
Afterwards, simulation was also considered using the stiction
model in the system to substantiate the proper functioning of
the developed algorithm.
3.2 Performance Assessment of MATLAB model (with
stiction block)
Figure 4. Plot of process variable vs. time along with valve
stiction to the previous SIMULINK model.
Figure 4 shows the generated data with number of data point.
Stiction model was used in the simulated simulink process,
which caused the process variable to oscillate. Above figure
demonstrate that oscillation pattern of the process clearly.
3.2.1 Results
Results fund form Pindex and FCOR algorithm are shown
below-
Pindex=0.1500
FCOR=0.1231
Performance index is found from study according to
minimum variance control law. This time it is also evident
that the control loop is performing poorly. As performance
index of both loop is less than 0.5 or close to zero. Both
Pindex and FCOR algorithm show similar results with the
assurance of the proper functioning of the developed
algorithm.
3.3 Performance Assessment of ASPEN HYSIS
Simulated process
Simulation process of a distillation column was
demonstrated by HYSIS modeling, it was about an overhead
circuit for a distillation column using more detail than just
the condenser module. Column pressure control was
achieved primarily by bypassing some of the overhead vapor
around the condenser. PIC-101 was used to control the
overhead accumulator pressure using the bypass. The top
tray pressure was controlled by PIC-100, which essentially
maintains a constant pressure drop between the top tray and
the overhead accumulator. If the system overpressures, PIC-
102 vents to flare. The reflux rate was on flow control, with
the overhead accumulator level being controlled by the
product rate. There was a temperature control at the bottom
of the column, and level control for the sump.
9

www.ijsr.net
3.3.1 Performance Assessment of Different Loops
Figure 5 shows different process data generated by the
simulation developed in Aspen HYSIS. PIC-100 controlled
the distillate flow rate. Oscillation was added to this
controller by a transfer function block. This stream was
cooled by a heat exchanger and then feed to the condenser.
From graph and calculated index it is clearly evident that the
loop was performing very badly. PIC-101 also controlled a
part of distillate flow stream. LIC-100 controlled the level of
condenser. Valve stiction was added to this controller. FIC-
100 controls the reflux ratio of the distillation column.
Though there were some noise in the loop but the controller
is performing satisfactorily. TIC-100 controlled the heat duty
of the reboiler. This loop was very sensitive as this
controlled the different temperature of the distillation
column. From result it is evident that controller is
performing according to desire. Level indicator controller
named by LIC-101 was also performing well.
Figure 5. Plot of Controlled Variable Data of Simulated
distillation column by HYSIS.
3.3.2 Results
According to minimum variance control law, table 1 shows
that control loops FIC-100, LIC-101, LIC-102, PIC-101, and
TIC-100 are performing satisfactorily as their performance
indexes are closed to 1 for the considered distillation
column. But performance index of PIC-100 loop is very
much less than 0.5. This indicates that the controller is
performing badly. This is the causal outcomes, as at this
control loop oscillation was introduced while simulation was
performed. This is a significant demonstration of the
developed way that it will work relevantly even in real
process with the existing oscillations in system.
Table 1. Results showing performance index of different
control loops of HYSIS simulation.
Above table clearly shows the compatibility of the FCOR
algorithm for any real process simulation with oscillation
and noise inherent with the process systems.
3.4 Process Data Analysis of SAFCO Plant
Process data from SAFCO was tested against both of the
control loop performance assessment algorithms. Figure 6
shows 21FIC105.PV control loop data for a flow control
loop as an example of several loops in that fertilizer plant.
Figure 6. Plot of process variable vs time
3.4.1 Results
It can be seen from the table 2 is that there are 3 flow
indicator controllers and rest of the controllers are mainly
pressure indicator controller. If all the controllers are
analyzed then it can be seen that flow indicator controller is
performing well than the pressure indicator controller in
SAFCO. It can also be seen that pindex result shows that
some of the index results are very close to zero which means
the controller is performing very poorly. This completely
supports the developed algorithm, which strengthens the
pragmatic convenient use for any real process plant like
fertilizer cpmplex.
Table 2. Results showing performance index of different
control loops of the SAFCO plant
10

www.ijsr.net
4. Outcomes and Findings
This project is basically based on closed loop univariate
system for single input and single output (SISO) process to
find out whether is the existing controller is working
satisfactory or not. By comparing the performance index of
the target loops with the theoretical standard the plant
operator can identify the faulty loops only by analysis the
routine operating data. Future work could be extended to
multivariate system for multiple inputs and multiple outputs
(MIMO) process. Topics of further research adjunct to
adjoincy matix with Minimum variance control, Controller
Auto-Tuning Based on Control Performance Monitoring,
Online performance assessment, Automation of the
controller diagnosis. By only analyzing closed loop data,
unsatisfactory loops can be identified easily from thousands
of control loops. In modern process systems there are strong
incentives for automated control performance monitoring
(CPM) and it’s assessments for process control loops or the
controller. Although several CPM techniques have been
applied successfully already, they also have several
shortcomings. First, most of the existing techniques assess
control system performance but do not diagnose the root
cause of the poor performance. A second shortcoming is that
assessment formulation for MIMO process is complicated
and usually restricted to unconstrained control systems,
monitoring strategies for MPC systems are a subject of
current research all over the world.
References
[1]"Shoukat Choudhury, M. A. A; Thornhill, N.F.;
Shah, S.L.;", Modelling valve stiction, Control
Engineering Practice, 13, 5, 641-658, 2005,
Elsevier
[2]"Ljung, L.; ",System identification - Theory for the
User,1999,Prentice-Hall
[3]"Huang, B.; Kadali, R.;","Dynamic modeling,
predictive control and performance monitoring: A
data-driven subspace approach",374,2008,Springer
Verlag
[4]“Manu Huttunen; Olli Pyrhnen; Riku Pllnenc; Ari
Koponen;” Real-time simulation environment for
control loop performance monitoring using matlab.
2012
[5]"Desborough, L ;Harris,T ;" "Performance
Assessment Measure for Univariate Feedback
Control" Canadian Journal of Chemical
Engineering, 70, 1186-1197, 1992
11

Performance assessment of control loops

More Related Content

What's hot (19)

Similar to Performance assessment of control loops (20)

More from International Journal of Science and Research (IJSR) (20)

Recently uploaded (20)

Performance assessment of control loops