[ieee 2012 ieee international conference on control system, computing and engineering (iccsce) -...
TRANSCRIPT
Reformulated Tangent Method of Various PID Controller Tuning for Air Pressure Control
Najidah Hambali, Mohd Nazmi Kamarol Zaki
Faculty of Electrical Engineering Universiti Teknologi MARA Malaysia 40450 Shah Alam, Selangor, Malaysia
[email protected] [email protected]
Abdul Aziz Ishak Faculty of Chemical Engineering
Universiti Teknologi MARA Malaysia 40450 Shah Alam, Selangor, Malaysia
[email protected] http://aabi.tripod.com
Abstract—The optimum proportional(P), integral(I) and derivative(D) constants can be estimated using tangent method and proper tuning rule for process stability. The purpose of this paper is to compare the performance of various controller tuning methods for controlling the process of air pressure control trainer model SOLTEQ SE121. The simplest open loop test on actual plant has been performed using a new Reformulated Tangent Method (RTM). The optimum PID value is set to the real plant based on Ziegler-Nichols (ZN), Chien-Hrones-Reswick (CHR) and Chien et. al. (Chien) tuning rules. The PID Tuner software is used to simulate the response to be compared with the real plant response. The plant is connected to a Yokogawa Centum CS 3000 R3 Distributed Control System (DCS). The performance of each method is analyzed in term of percent overshoot(%OS), settling time(Ts), rise time(Tr) and integral absolute error(IAE).
Keywords: Reformulated Tangent Method, Controller Tuning, optimum PID, Air Pressure Control Trainer, Distributed Control System
I. INTRODUCTION
More than 97% of the regulatory controllers in process industries are of the PID type since early 1940s. Begin with pneumatic control; the PID controllers have survived in many changes in technology through direct digital control to the DCS.
PID control is fundamental control technology and it makes up 90% of automatic controllers on process control fields [1]. It is very important to keep the process working probably and safety in the industry for the quality of the product being processed, environment issues and for total energy saving system and thus the PID control is absolutely essential. In 1943, ZN method which is for first-order-plus-time-delay (FOPTD) was proposed by Ziegler and Nichols which target 25% Quarter Amplitude Decay (QAD) [2].
There are many suggestions for modifications of the Ziegler-Nichols methods. One of them is the Chien, Hrones and Reswick method (CHR). CHR method gives closed loop systems with slightly better robustness than the Ziegler-Nichols method. The design used was “quickest response without overshoot” [3]. Chien method is also based on the FOPTD process model. Chien tuning for set-point response type gives closed loop system with 20% QAD and minimum response time [4].
This paper presents a performance comparison of 3 classical PID tuning methods ZN, CHR, Chien based on percent overshoot (%OS), settling time (Ts), rise time (Tr) and integral absolute error (IAE). In this work it is aim to tune the PID controller parameter to achieve best control performance. The calculation of Ts, Tr and %OS based on exact formula can affect the result of performance of plant [5]. Realizing the result of PID always not satisfying, there are practical controllers tuning tips that can make controller tuning task more efficient [6]. Ziegler-Nichols [2], Cohen Coon [7], Ogunnaike[8] and Seborg [9] proposed the open loop time response techniques to get the controller's gains.
Nowadays lot of researchers tries to get the best performance of their plant by comparing various techniques for example the Glycerin Bleaching Process [10]. Currently researchers try to develop the auto tuning PID controllers using Graphical User Interface (GUI) [11]. Other than that, the usage of Fuzzy Logic controller to improve the performance of plant is widely researched [12]. PID application not only used to control pressure plant but also can control water level and temperature plant [13, 14].
A new proposed of tangent method technique transforms the process response rate calculation into a trigonometric function to get the optimum PID [15, 16]. The Reformulated tangent method simplifies the conventional tangent method analysis of process characteristics of the open loop response curve. It reduced the steps in analyzing the process response rate and improved data extraction speed.
This paper is organized as follows. In Section II, a brief description of the air pressure control trainer plant is performed. Section III describes the open loop test that has been used, before implementing Reformulated Tangent Method of various tuning rules. The experimental work is discussed in Section IV. In Section V, simulation studies, the results of open loop test and closed loop set point change are presented. Conclusions are made in Section VI.
II. AIR PRESSURE CONTROL TRAINER PLANT
This plant which installed in Process Control Laboratory, Faculty of Electrical Engineering, Universiti Teknologi Mara(UiTM), Shah Alam is a standard
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
978-1-4673-3143-2/12/$31.00 ©2012 IEEE 17
industrial pressure control loop that hasprovide student with hands-on experpressure loops can be controlled using based controller as in Figure 1. The Trainer is an air process where 6 bar of charged into the air receiver tank approximately 6 bar, by an air regulator.tank is passed through the process linecontrol tank, which is filled to read 3 meter read 5 m3/hr by regulating hand va
Figure 1: The Air Pressure Control Tra
A pressure transmitter is installed
control tank and its output linked to controller. The output from the controcontrol valve which is installed at the controller regulates the control valve opto maintain the pressure inside the ccontrol panel is connected to a Yokog3000 R3 Distributed Control System (three standards required component fHuman Interface Station (HIS) in CS which the operator controls the plancomponent can also be used to do confiThe DCS consists of the Field Contrwhich is the interface between the field icontrol room. The FCS and HIS are cotime control network called thecommunicates all the parameters to anHIS.
III. OPEN-LOOP TES
In order to perform an open loop tes
been stabilized either in automatic modeA step change of 5 to 20% has been permode. Test completed when process reastate process variable (PVnss) for self rBased on open loop test, a process caself-regulating, integrating, or runaway [
The three controller parameters can performing an open loop or some dynactual plant. The simplest test that can bReformulated Tangent Method [15, 16the output response of the open loopprocess.
s been designed to rience on how a a microprocessor-Pressure Control
f compressed air is and regulated to . Air from receiver e into the pressure
bar and set flow alve.
ainer Plant
d at the pressure a microprocessor
oller is sent to a process line. The
pening for airflow ontrol tanks. The gawa Centum CS (DCS). There are for the DCS The
3000 R3 system nt and the same
figuration changes. rol Station (FCS) instrument and the
onnected via a real e V-Net which nd from FCS and
ST
st, the process has e or manual mode. rformed in manual aches a new steady egulating process.
an be classified as [17]. be determined by
namic test on the be performed is the ]. Figure 2 shows p test for typical
Figure 2. Transforming process rat
Response rate, RR;
ba
ΔMVtanθRR =
Dead time, Td;
( −= PVPVT osnssd
Time constant, Tc;
⎢⎣
⎡ −=
tanθPVPV
abT nss
c
Where: RR : Response Rate, 1/tim ΔMV : change in controller’ a : conversion factor of b : conversion factor of
IV. PID
The general PID algorithm
CS3000 as in (14);
⎢⎣⎡ += ∫I
1E(t)P
100MV
Since in this paper the controalgorithm form is:
⎢⎣⎡ += ∫I
1y(t)P
100MV
Where: MV : Manipulated VariablE(t) : Error y(t) : Process Variable P : Proportional Band (RI : Reset Time, Second
Seconds) D : Derivative Time, Sec
Seconds)
te into trigonometric form
(1)
) ⎥⎦
⎤⎢⎣
⎡−⎟
⎠⎞
⎜⎝⎛
tanθ1
tanβ1
ab
s (2)
⎥⎦
⎤Voss (3)
me s output,% PV, %/length time, time/length
TUNING
ms form of the Yokogawa
⎥⎦⎤+
dtdE(t)DE(t)dt (4)
oller type is I-PD, then the
⎥⎦⎤E(t)dt (5)
le
Range: 0 to 1000) (Range 0.1 to 10,000
conds (Range 0-10,000
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
18
There are three different controller calculation algorithms associated with the PID Block as shown in Table 1.
TABLE 1: CONTROLLER CALCULATION ALGORITHMS
PID Control Algorithm
Calculation Input VariableProportional
Term Integral
Term Derivative
Term PID On Error On Error On Error
I-PD On Measurement On Error On
Measurement
PI-D On Error On Error On Measurement
Automatic Determination
Same as I-PD in the AUT ModeSame as PI-D in the CAS or RCAS Mode
Automatic Determination 2
Same as I-PD in the AUT or RCAS ModeSame as PI-D in the CAS Mode
The default algorithm is Automatic Determination 2. The combination of PB and I value will determine whether the process is oscillatory, stable or unstable. There are three open loop tuning rules that use set point change, which have been compiled by Senbon and Hanabuchi [15]. The formula of the tuning method can be found in Table 2.
TABLE 2: VARIOUS METHODS OF CONTROLLER TUNING SETTING
Tuning Method PB (%) I (s)
ZN 111.1RRTd 3.33Td
CHR 286RRTd 1.2Tc
Chien 167RRTd Tc
V. EXPERIMENTAL WORK
The experimental work procedure is shown in Figure 3. Firstly the open loop test is conducted to real pressure plant. After that by using the Reformulated Tangent Method, the value of RR, Td and Tc is calculated to determine the PB and I based on tuning rules. Then, the DCS is operated in closed loop or automatic mode to proceed with the set point change operation. The loop performance in terms of Tr, Ts, %OS and IAE is analyzed. Finally, the fine tuning technique is implemented to the best tuning method performance which has been selected in terms of Tr, Ts, %OS and IAE.
Figure 3: Flow chart for experimental work
VI. RESULTS AND DISCUSSION
The open loop response curve shown in Figure 4 is a self-regulating process. The value of process parameters after calculation using Reformulated Tangent Method are Td = 4.28 s, Tc= 32.15 s and RR = 0.029. The value of scale for y-axis is % and x-axis is .
Figure 4: Open loop test and Reformulated Tangent Method
θ
Time,(second)
PV ,(
%)
Td Tc
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
19
A control without D mode is used of the system is not required. The valueand I using three tuning rules is shown in
TABLE 3: PROPORTIONAL BAND AND INTEGR
Tuning Method PB (%)
ZN 13.78
CHR 35.49
Chien 20.73
Figure 5, Figure 6 and Figure 7 shcurves after using various tuning meTuner software.
Figure 5: Simulation of ZN m
Figure 6: Simulation of CHR met
Figure 7: Simulation of Chien me
Figure 8, Figure 9 and Figure 10 show curves for ZN, CHR and Chien reresponse curves from simulation and Dsimilar. From Figure 8, the ZN and Chiedid not achieved the settling criteria 25respectively. However, CHR method achcriteria with no overshoot but not acresponse time. All tuning method nsettling criteria because the algorithm usI-PD algorithm. In this system, the pr
when fast respond e of calculated PB n Table 3.
RAL TIME VALUE
I (s)
14.25
38.59
32.2
how the response ethods using PID
method
thod
ethod
the DCS response espectively. Both CS techniques are en response curves % and 20% QAD hieved the settling chieved minimum not achieved their sed in this study is roportional action
cannot cause an abrupt changea stable control characteristic c
Figure 8: ZN Res
Figure 9: CHR Re
Figure 10: Chien R
It is noted that in time reCHR response was achieved wtimes, and IAE of 72.85 respectively. While for the Chsettling times, and IAE of respectively. Amongst all the had shown the best rise and 27.85 s, 60 s and 2.77. corresponding the rise and setttuning method. A faster respoachieved as compared to the CH
e of manipulated output, but an be obtained.
sponse Curve
esponse Curve
Response Curve
esponse specifications, the within the rise and settling s, 137.14 s and 8.0625 hien response, the rise and 75 s, 113.75 s and 5.17 methods, the ZN response settling times, and IAE of Table 4 summarizes the tling times, and IAE for all onse for ZN controller was HR and Chien controller.
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
20
TABLE 4: CONTROL PERFORMANCE FOR DIFFEREN
Tuning Methods Tr (s) Ts (s) %
ZN 27.85 60.00
CHR 72.85 137.14
Chien 75.00 113.75
There are many ways to improve thetuning technique is used in this study, bI not more than twice, while decrease PBhalf to get better result. The fine tunincurve is shown in Figure 11 while Tablecorresponding the rise and settling timefor ZN before and after fine tuning.
Figure 11: Fine Tuning on ZN Out
After the fine tuning process with PB of14.25 s, ZN response had better rise 6.42 s and 2.04 respectively. It is noteresponse achieve settling time at 9015.71 % of OS.
TABLE 5: DIFFERENCE PERFORMANCE OFBEFORE AND AFTER FINE TUNI
Fine Tuning PB (%) I (s) Tr (s) T
Before 13.78 14.25 27.85 60
After 27.58 7.13 6.42 90
The ZN open loop tuning rule has sThe ZN rule is very sensitive to an accuof dead time, which is difficult on lag-dowith short dead times.
VII. CONCLUSION
Analysis of the performance comPID tuning method using Reformulatedhave been presented. Performances of tChien controller are examined in termsspecifications. The results had shown thwas the best tuning method for the airtrainer model SOLTEQ SE121. The tun
NT TUNING METHODS
% OS IAE
- 2.77
- 8.0625
- 5.17
e result. The fine y increase PB and
B or I not less than ng of ZN response e 5 summarizes the es, % OS and IAE
tput Response
f 13.78 % and I of time and IAE of
ed that, the output 0 s and produce
F ZN METHOD ING
Ts (s) %OS IAE
0.00 - 2.77
0.00 15.71 2.04
several drawbacks. urate measurement ominant processes
mparison for three d Tangent Method the ZN, CHR and s of time response hat the ZN method r pressure control ning method used
in this paper is classical methothis research is by applying othCohen-Coon and Takahashi. various open loop response cufind best process parameter comparison between PID algormade to give best performancethat, other set point value can bof tuning method. Since this sdisturbance, hence the usage ocan also be implemented.
ACKNOWLE
The work is financed Mara(UiTM), Shah Alam, Electrical Engineering.
REFEREN
[1] Carl Knospe, “PID Contropp.30-31, 2006.
[2] J. G. Ziegler, N. B. NichAutomatic Controllers,” Tra1942.
[3] K. J. Astrom, T. HagglunInstrument Society of Amer
[4] Aidan O’Dwyer, “HandbooTuning Rules,” 3rd Ed., Imp2009.
[5] Norman S. Nise, “Control John Wiley & Son Inc., pp. 1
[6] Armando B. Corripio, “TSystems”, 2nd Ed., The IAutomation Society, pp. 74-
[7] G.H. Cohen, and Coon, G. ARetarded Control”, Trans. A1953
[8] B. Ogunnaike and W.H.Modeling and Control”, Oxf
[9] D.E. Seborg, T.F. Edgar, anDynamics and Control”, Engineering, 1994.
[10] N. Kamaruddin , Z. Janin, Controller Tuning for GlycWell-Known Tuning Formu35th Annual Conference of I
[11] T.L. Mohamed, Z. MohamTuning PID Controller usi(GUI)”, 2nd InternationalEngineering and Application
[12] Z. Yusuf, Z. Janin, M. N. Tfor Glycerin Bleaching International Colloquium oApplication, pp. 250-254, 20
[13] S. Isa, B. C. C. Meng, Z. SaStudy of PID Controlled Level Measurement SysteColloquium on Signal Proce237-242, 2011.
ods. Hence, future work for hers tuning methods such as
Moreover comparison on urve calculation methods to can be used. Furthermore rithms structure also can be e to the process. Other than be made to see performance study used set point change of change load disturbance
DGMENT by Universiti Teknologi
supported by Faculty of
NCES ol”, IEEE Control Magazine,
hols, “Optimum Settings for ans. ASME, 64, pp. 759-768,
nd, “Advanced PID Control”, rica, pp. 166-167, 2006.
ok of PI And PID Controller erial College Press, pp. 30-31,
System Engineering”, 6th Ed, 178-181, 2010. Tuning of Industrial Control Instrumentation, System, and 76, 2001.
A., “Theorical Consideration of ASME, Vol. 75 pp. 827-834,
. Ray, “Process Dynamics, ford University, 1994 nd D.A. Mellichamp, “Process
Wiley Series in Chemical
Z. Yusuf, M. N. Taib, “PID erin Bleaching Process Using las-A Simulation Study”, 2009 IEEE, pp. 1682-1686, 2009. med, “Development of Auto ing Graphical User Interface l Conference on Computer n, pp. 491-494,2010. Taib, “Fuzzy Logic Controller
Temperature Control”, 6th on Signal Processing & Its 010. aad, N. A. Fauzi, “Comparative Modes On Automatic Water
em,” IEEE 7th International essing and Its Applications, pp.
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
21
[14] N. H. A. Hamid, M. M. Kamal, F. H. Yahaya, “Application of PID Controller in Controlling Refrigerator Temperature,” 5th International Colloqulum on Signal Processing & Its Application,pp. 378-384, 2009.
[15] A. A. Ishak, “Reformulation of the Tangent Method for PID Controller Tuning”, TENCON IEEE Region, Kuala Lumpur, pp. 1-5, 24-27 September 2000.
[16] A. A. Ishak, “Open-loop Process Identification: Reformulation of Response Rate Calculation”, Regional
Symposium on Chemical Engineering Thailand, pp. 484 – 488, 1998.
[17] J. F. Smuts, “Process Control for Practitioners”, OptiControl Inc., pp. 27-30, 2011.
[18] R. C. Rice, “PID Tuning Guide for The Yokogawa CS3000 and Centum VP DCS”, Control Station Inc, pp. 24-25, 2010.
[19] T. Senbon, F. Hanabuchi, “Instrumentation System: Fundamentals and Application”, Springer Verlag, pp. 49-50,1991.
2012 IEEE International Conference on Control System, Computing and Engineering, 23 - 25 Nov. 2012, Penang, Malaysia
22