Você está na página 1de 21

A Review of Project Paper #24

A simple method of tuning PID controllers for stable and unstable FOPTD systems (R. Padma Sree, et al.)
MSE 490 Advanced PID Control Kjell Sadowski 12/04/2014

Abstract This paper reviews the PI/PID controller design method proposed by R.Padma Sree, M. N. Srinivas and M. Chidambaram in their paper titled: A simple method of tuning PID controllers for stable and unstable FOPTD systems. A derivation of the theory for the proposed method is displayed and further analyzed. Results of the relevant sections are replicated using simulations in MATLAB to check the correspondence of them to the results they are reporting. Further analysis will show the discrepancies between results from the simulations and the ones reported in the paper. A discussion of the results will be presented along with a brief literature review.

CONTENTS
List of Figures...................................................................................................................................................................... ii List of Tables ....................................................................................................................................................................... ii Glossary ................................................................................................................................................................................ ii 1 2 3 Introduction ................................................................................................................................................................ 1 Problem Statement .................................................................................................................................................... 1 Theory.......................................................................................................................................................................... 2 3.1 3.2 3.3 4 4.1 4.2 5 6 7 8 9 Derivation of Proposed Method .................................................................................................................... 2 Derivation of Two-Parameter PID Tuning Method .................................................................................. 5 Analysis of Proposed Method ........................................................................................................................ 6 PI Controller ..................................................................................................................................................... 7 PID Controller ................................................................................................................................................ 10

Simulation Studies...................................................................................................................................................... 7

Analysis of Results ................................................................................................................................................... 14 Discussion ................................................................................................................................................................. 16 Literature survey ...................................................................................................................................................... 16 Conclusion ................................................................................................................................................................ 16 Bibliography .............................................................................................................................................................. 17

Appendix A: MATLAB Error Calculation Code......................................................................................................... 18

Page i

LIST OF FIGURES
Figure 1 - PI Controller Set Point Response .................................................................................................................. 8 Figure 2 - PI Controller Regulatory Response ............................................................................................................... 8 Figure 3 - PI Controller Uncertainty in Parameters Regulatory Response ................................................................ 9 Figure 4 - PI Controller Uncertainty in Parameters Set Point Response................................................................. 10 Figure 5 - PID Controller Set Point Response ............................................................................................................ 11 Figure 6 - PID Controller Regulatory Response.......................................................................................................... 12 Figure 7 - PID Controller Parameter Uncertainty Regulatory Response ................................................................ 13 Figure 8 - PID Controller Parameter Uncertainty Set Point Response ................................................................... 14

LIST OF TABLES
Table 1 - Analysis of System Types ................................................................................................................................. 6 Table 2 - PI Controller Errors .......................................................................................................................................... 7 Table 3 - PI Controller Parameter Uncertainty Errors ................................................................................................. 9 Table 4 - PID Controller Errors..................................................................................................................................... 11 Table 5 - PID Controller Parameter Uncertainty Errors ........................................................................................... 12 Table 6 - PI Controller Simulation Results Comparison ............................................................................................ 14 Table 7 - PI Controller Parameter Uncertainty Simulation Results Comparison................................................... 15 Table 8 - PID Controller Simulation Results Comparison ........................................................................................ 15 Table 9 - PID Controller Parameter Uncertainty Simulation Results Comparison ............................................... 15

GLOSSARY
FOPTD First Order Plus Time Delay IAE Integral Absolute Error IMC Internal Model Control ISE Integral Square Error ITAE Integral Time Absolute Error PID Proportional Integral Derivative SOPTD Second Order Plus Time Delay Z-N Ziegler-Nichols

Page ii

1 INTRODUCTION
The main idea behind the development of this method was to design a tuning method for a FOPTD model which was simple and could be applied to both stable and unstable systems, which would be designed for set point tracking and would provide an overall good response. This method would be able to be used to calculate tuning parameters for both PI and PID type controllers. In this report, the proposed tuning method of the analyzed paper [1] and the results for a stable FOPTD PI and PID controller will be reproduced and analyzed along with the results from the related simulations. Although this paper provides an analysis and results for unstable systems, this data will not be presented because it is not within the scope of this analysis.

2 PROBLEM STATEMENT
The tuning method outlined in this paper was design for closed loop control of a set point tracking problem or formally known as the Servo problem. The authors report that the simulation results show that the method gives a similar response as that of Ziegler-Nichols (Z-N) method and better response than that of IMC method [1]. The assumptions that the authors make when deriving the method for determining the controller parameters are: The system can be forced to reach the set point: () () = 1, and The transfer function is open-loop stable. The approach which the authors of this paper have taken to calculate the PID parameters is very different than what many other control systems researchers have derived in the past. Compared to a method like the Ziegler Nichols Continuous Cycling Method [2] which calculates the parameters from the actual system, these parameters rely solely on the FOPTD model being as accurate as possible. The controller parameters calculated from this method will be very sensitive to small changes in the system, which would make this method difficult to implement in systems that have very dynamic system states.

Page 1

3 THEORY
This section shows a complete derivation of the PID parameter equations presented in the analyzed paper [1], a derivation of the two-parameter tuning equations and an analysis of the equations for determining the PID parameters.

3.1 DERIVATION OF PROPOSED METHOD


This section shows the derivation of the proposed method for the PID Controller parameter equations which are presented in the analyzed paper [1]. Equations (1) to (14) in the analyzed paper are shown here and do not correspond to the numbers attached to the equations in this section. : () = :
+1

() 1 = [1 + + ] () 1

Closed Loop Transfer Function: (1 + () = 1 + ) 1 (1)

1 (1 + + ) 1 + 1] ( + 1) [ ( + 1)

1 (1 + + ) 1 =( )( ) 1 (1 + + ) + ( + 1) 1 + 2 = [ ( + + 2 ) + 2 + + Setting a change of variables: = =


(2)

(3)

2 + + 2 0.5 = [ ( ) ] 0.5 2 2 ( + + 2 ) + +
[

(4)

2 + 2 ( + + ) + ( + 1) +

(5)

Page 2

Making the following variable substitutions 1 = 2 = 3 = = 1 1

The term is removed from the numerator of equation (5) and separated from the transfer function because this function only provides a time shift. The closed loop transfer function of an open loop stable transfer function is analyzed starting with equation (6). (1 + 2 + 3 2 ) 0.5 () = [ ] (1 + 2 + 3 2 ) 0.5 + ( + 1) 0.5 () Taylor series expansions 0.5 = 1 + 0.5 + 0.125 2 2 + 0.02083333 ^3 0.5 = 1 0.5 + 0.125 2 2 0.02083333 ^3 (7) (8) (6)

Since the objective of the transfer function is to reach the point where the system reaches the set point and settles there, () () = 1 will be sets and the analysis will be carried using this result and numerator and denominator of equation (9) will be equated. (1 + 2 + 3 2 ) 0.5 () = 1 = [ ] (1 + 2 + 3 2 ) 0.5 + ( + 1) 0.5 () ( 2 + ) 0.5 + (1 + 2 + 3 2 ) 0.5 = (1 + 2 + 3 2 ) 0.5 2 ( 0.5 + 3 0.5 ) + ( 0.5 + 1 0.5 ) + 2 0.5 = 2 (3 0.5 ) + (1 0.5 ) + +2 0.5 (0.5 + 0.125 2 0.53 + 0.125 2 1 0.0208333 3 2 ) 3 + (0.5 + 3 0.51 + 0.125 2 2 + 1) 2 + (1 0.52 + 1) + 2 = (0.53 + 0.125 2 1 + 0.0208333 3 2 )3 + (3 + 0.51 + 0.125 2 2 )2 + (1 + 0.52 ) + 2 1 : 1 0.52 + 1 = 1 + 0.52 1 = 2 2 = 1 (13) (9) (10) (11)

(12)

Page 3

2 :

0.5 + 3 0.51 + 0.125 2 2 + 1 = 3 + 0.51 + 0.125 2 2 0.5 0.51 + 1 = 0.51 0.5 + 1 = 1 1 = 1 + 0.5 (14)

3 :

0.5 + 0.125 2 0.53 + 0.125 2 1 0.0208333 3 2 = 0.53 + 0.125 2 1 + 0.0208333 3 2 1 1 0.5 + 0.125 2 0.0208333 3 ( ) = 0.53 + 0.0208333 3 ( ) 0.5 + 0.08333 2 = 3 3 = 0.5 + 0.08333 1 = = = 1 + 0.5 (16) (15)

+ 0.5 1 1 = =

2 =

1 = ( + 0.5) = + 0.5 3 = 0.5 + 0.08333 ( ) 1 = 1 = 0.5 + 0.08333


2 0.5 + 0.08333 0.5 + 0.08333 ( ) = + 0.5 + 0.5

(17)

0.5( + 0.1667 ) + 0.5

(18)

Equations (16), (17) and (18) describe the parameters for the system gain, reset time and derivative time. If the time delay ( ) and time constant () are known, all of these parameters can be calculated.

Page 4

3.2 DERIVATION OF TWO-PARAMETER PID TUNING METHOD


This section shows the derivation of the two-parameter controller optimization method presented in the paper of analysis [1]. This section describes the development of equation (17) in the studied paper. Starting with equation (12), the left-hand side is multiplied by a constant, #. q1 is multiplied by 1 and q2 and q3 are multiplied by 2, which has a linear relation to 1 defined by the relation: 2 = 1 The expansion of equation (12) follows (19)

(0.5 + 0.125 2 0.53 + 0.125 2 1 0.0208333 3 2 )2 3 + (0.5 + 3 0.51 + 0.125 2 2 + 1)2 2 + (1 0.52 + 1)1 + 2 = (0.53 + 0.125 2 1 + 0.0208333 3 2 ) 3 + (3 + 0.51 + 0.125 2 2 )2 + (1 + 0.52 ) + 2 1 (1 0.52 + 1) = 1 + 0.52 1 1 0.52 1 + 1 = 1 + 0.52 = 1 1 + 0.52 + 0.52 = (1 1 )1 + (1 + 1 )0.52 2 (0.5 + 3 0.51 + 0.125 2 2 + 1) = (3 + 0.51 + 0.125 2 2 ) 2 (1 0.5) = 3 2 3 + 0.51 + 0.51 2 + 0.125 2 2 0.125 2 2 2 2 (1 0.5) = (1 2 )3 + 0.5(1 + 2 )1 + 0.125 2 (1 2 )2 1 0.25 3 ( ) (0.5 + 0.125 2 0.53 + 0.125 2 1 2 ) 2 0.5 12 0.25 3 1 = (0.53 + 0.125 2 1 + 2 ) ( ) 12 0.5 2 + 0.25 3 + 0.251 2 0.25 2 2 2 = 3 + 0.251 6 0.25 2 2 2 6

(20)

(21)

(22)

(1 + 0.25) = 3 (1 + 2 ) + 0.251 (1 2 ) +

(23)

Page 5

Using equations (21), (22) and (23) to form a matrix of the linear equations: 0.5(1 + 1 ) 0 1 2 (1 0.125 2 ) 1 2 1 (1 [ 2 ] = [ 2 0.5) ] 0.25 2 (1 0.25) (1 + 2 ) 1 + 2 ] 3 [0.25(1 2 ) 6 1 1 0.5(1 + 2 )

(24)

As stated in the analyzed paper [1], = 0.6 for the PI controller equations and 0.8 for the PID controller equations. Insufficient parameters are presented in the paper to calculate the equations for the PI controller so only the equations will be presented. = 0.9719 0.8915 = 0.7719 4 3.6608 3 + 6.5791 2 + 5.1652 + 2.8059 (25) (26)

For the PID controller design, = 0.8 as stated above, 1 = 1.2 and 2 is calculated using equation (19). The value corresponds to the model and is approximately but not exactly 0.5. The equations for the PID controller developed using this method are: = 1.377 0.8422 = 1.085 0.4777 = 0.3899 + 0.0195 (27) (28)

(29)

3.3 ANALYSIS OF PROPOSED METHOD


In order to understand the effect of varying time delay and time constant on equations (16), (17) and (18), the limits of the gains have been taken and the results are summarized in Table 1.
Table 1 - Analysis of System Types

System Type Very slow Very fast Large time delay Short time delay

Limit lim (, )
0 0

System Gain = 0.5 = = 0.5 =

Integral Time = 0.5 = = =

Derivative Time = 1 = /6 = = 1

lim (, ) lim (, ) lim (, )

In the case where a system is very slow, the controller uses small gains to control the system. Derivative action is the same as the system gain and the integral gain is the inverse of the time delay. So a slow system with a small time delay will be mainly controlled by the integral action and a slow system with a large time delay will use a very sluggish controller, where the derivative action will be the driving force. Page 6

A very fast system will not require a PID controller according to these tuning rules. The integral action becomes zero and the derivative action becomes very small compared to the system gain so control using the difference in error, or proportional action, will be enough to control the system. A system with a very large time delay will be controlled by a PD controller where the derivative action is the main driving component of the controller. A system with a very short time delay relies mainly on the system gain and derivative action. Integral action may also play a major part in the system action if the system has a very small time constant. It is also interesting to note that the analyzed paper [1] makes no mention of using a filter on the derivative of the controller; thus, all simulation studies will be formed without using one for the proposed method.

4 SIMULATION STUDIES
This section shows all of the replicated results from the paper of study [1]. All results were replicated using MATLAB to be as similar as possible to the original figures given that enough information was provided in the section descriptions. Comments regarding major deviations in the results will be made in Section 5.

4.1 PI CONTROLLER
The simulation results of the PI controller development using equations (25) and (26), which were partially developed in Section 3.2, are shown in this section. Graphs for both the set point response and the regulatory response are shown. Graphs for parameter uncertainty are also provided where each FOPTD model parameter is increased by 20% individually. Table 2 provides the information for the errors and Figure 1 and Figure 2 show the results of the proposed method compared with the Ziegler-Nichols [3] and Abbas [4] PI controller development methods. Table 3 provides ISE error data for the responses to parameter variations. Figure 3 and Figure 4 are the graphs which describe these responses.
Table 2 - PI Controller Errors

Method Present Z-N Abbas

Servo Problem ISE IAE 1.4227 2.0192 1.7699 3.6769 2.1632 3.5847

ITAE 3.1423 17.7768 10.9859

Regulatory Problem ISE IAE 0.3576 0.8125 0.3070 0.9335 0.7102 1.9993

ITAE 1.3198 3.4799 7.4831

Page 7

Figure 1 - PI Controller Set Point Response

Figure 2 - PI Controller Regulatory Response

Page 8

Table 3 - PI Controller Parameter Uncertainty Errors

Method Present Z-N Abbas

Servo Problem ISE values for uncertainty in 120% kp 120% 1.4458 1.4840 1.5626 1.8189 1.8911 2.2488

120% d 1.7610 1.8964 2.2849

Regulatory Problem ISE values for uncertainty in 120% kp 120% 0.4719 0.3176 0.4156 0.2621 0.8911 0.6653

120% d 0.3918 0.3288 0.7342

Figure 3 - PI Controller Uncertainty in Parameters Regulatory Response

Page 9

Figure 4 - PI Controller Uncertainty in Parameters Set Point Response

4.2 PID CONTROLLER


This section contains the simulation results for the PID controller using no tuning parameters and two tuning parameters, and comparing them against the results for the Ziegler-Nichols [3] and IMC [5] [6] methods. The graphs are based off of equations (16), (17) and (18) for the model with no tuning parameters and equations (27), (28) and (29) for the model with two tuning parameters. Graphs for parameter uncertainty are also provided where each FOPTD model parameter is increased by 20% individually. Table 4 provides the errors for the set point and regulatory responses of all four systems and Figure 5 and Figure 6 show the graphs of the responses. Table 5 provides the ISE error information for the set point tracking and regulatory responses with parameter variation; Figure 7 and Figure 8 show the graphical responses.

Page 10

Table 4 - PID Controller Errors

Servo Problem ISE IAE Presenta 0.6781 0.9691 b Present 0.7520 1.1165 Z-N 0.6939 1.0176 IMC 0.7146 0.9968 a No tuning parameters b Two tuning parameters

Method

ITAE 0.7913 0.9582 0.8876 0.7675

Regulatory Problem ISE IAE 0.0833 0.4247 0.0966 0.4076 0.1173 0.5112 0.1202 0.4754

ITAE 0.7433 0.5650 0.8511 0.6868

Figure 5 - PID Controller Set Point Response

Page 11

Figure 6 - PID Controller Regulatory Response

Table 5 - PID Controller Parameter Uncertainty Errors

Servo Problem ISE values for uncertainty in 120% kp 120% Presenta 0.7686 0.6860 Presentb 0.8252 0.7800 Z-N 0.7750 0.7126 IMC 0.7051 0.7591 a No tuning parameters b Two tuning parameters

Method

120% d 0.8961 1.0587 0.9070 0.8622

Regulatory Problem ISE values for uncertainty in 120% kp 120% 0.1217 0.0729 0.1141 0.0946 0.1504 0.1112 0.1453 0.1132

120% d 0.0897 0.1179 0.1217 0.1381

Page 12

Figure 7 - PID Controller Parameter Uncertainty Regulatory Response

Page 13

Figure 8 - PID Controller Parameter Uncertainty Set Point Response

5 ANALYSIS OF RESULTS
This section presents a comparison of the results of the simulations against those which were provided in the analyzed paper [1]. The discrepancies which exist are hypothesized to be due to the fact that the analyzed paper was written 10 years ago. This time discrepancy may allow for the development of better numerical methods used to evaluate the mapping of transfer functions. It is also worth noting that the authors of the paper did not provide any information about which mathematical function was used to test the disturbance rejection of the controller. A square impulse, ramp, impulse and sinc function were tried but only the ramp impulse provided a response that was stable as well as visually similar to those produced by the authors. Table 6, Table 7, Table 8 and Table 9 provide the results comparisons in relative percentages to the numbers presented in the analyzed paper.

Table 6 - PI Controller Simulation Results Comparison

Method Present Z-N Abbas

Servo Problem ISE -2% 0% 0%

IAE -1% 0% 0%

ITAE -4% -1% -1%

Regulatory Problem ISE IAE -51% -47% -75% -75% -56% -44%

ITAE -71% -86% -59% Page 14

Table 7 - PI Controller Parameter Uncertainty Simulation Results Comparison

Method Present Z-N Abbas

Servo Problem ISE values for uncertainty in 120% kp 120% -4% -1% -2% 0% 0% 0%

120% d -1% -2% -1%

Regulatory Problem ISE values for uncertainty in 120% kp 120% -51% -55% -71% -78% -54% -59%

120% d -57% -75% -58%

Table 8 - PID Controller Simulation Results Comparison

Servo Problem ISE Presenta 0% Presentb -1% Z-N 0% IMC -4% a No tuning parameters b Two tuning parameters

Method

IAE 1% 0% 0% -4%

ITAE 6% 0% 3% 3%

Regulatory Problem ISE IAE -24% -15% 0% 11% 16% 20% -31% -20%

ITAE -15% 12% 32% -28%

Table 9 - PID Controller Parameter Uncertainty Simulation Results Comparison

Servo Problem ISE values for uncertainty in 120% kp 120% Presenta 2% -1% Presentb 0% -1% Z-N 2% -1% IMC -4% -4% a No tuning parameters b Two tuning parameters

Method

120% d 0% 0% 0% -4%

Regulatory Problem ISE values for uncertainty in 120% kp 120% -8% -31% 20% -18% 23% 15% -27% -33%

120% d -35% 30% -5% -33%

Page 15

6 DISCUSSION
Aside from the issue regarding the disturbance rejection modelling described in Section 5, the results of the system appeared to look very good for PID controllers when compared to the Ziegler-Nichols and IMC models and for PI controllers when compared to Ziegler-Nichols and Abbas models. The results of the proposed method do exhibit some high overshoot but fast settling time also. It should be strongly noted that there are issues with the stability of PID (no tuning and two tuning parameter models) and PI controllers developed using this method. As seen in Figure 4, Figure 7, and Figure 8, these controllers do not display strong stability characteristics for parameter fluctuation; however, the disturbance rejection capability of PI controllers does not seem to be affected, as seen in Figure 3. In order to implement this algorithm, the following conditions should be met: 1. Adaptive control must be used, and 2. The system must not change faster than the adaptive control can recalculated the FOPTD model Adaptive control will ensure that the values which the controller are using are most appropriate for the current state of the system and thus ensure the best possible response. This type of method is suggested only for systems that can tolerate an underdamped response, require a quick settling time and have the capacity to use adaptive control. This method would not be appropriate for a system that has highly fluctuating parameters and does not need such a fast settling time. A more conservative and stable method such as IMC may be a better choice for such a system.

7 LITERATURE SURVEY
A survey of the papers provided for this project produced a paper which described the development of a simple and analytic controller tuning method. The paper titled Simple analytic rules for model reduction and PID controller tuning by S. Skogestad [7] describes a method which uses the IMC controller framework to create a simple set of tuning rules for FOPTD and SOPTD systems. These rules are simple equations where the input parameters are governed by relations which describe the types of output behaviors; for example, = 8 will produce a system that provides a good trade-off between disturbance response and robustness [7]. The advantage of this type of method is that the results for the model can be accurately predicted based on the type of system input. The disadvantage is that there are probably few real-life models which can satisfy such relations on a consistent basis. In comparison with the method in the analyzed paper [1], the method proposed by Skogestad provides more predictable results for more rigidly defined models, whereas the proposed method is more flexible and provides good results but the controller is not as robust.

8 CONCLUSION
This paper has reviewed the theory and simulation results depicted in the paper: A simple method of tuning PID controllers for stable and unstable FOPTD systems by R. Padma Sree, et al. A detailed derivation of the theory has been presented to show the steps required to reach the proposed equations along with an analysis of those equations. The models which were tested in the paper were reviewed and replicated using MATLAB simulation. The results from these simulations are presented with the discrepancies in the data. The set point Page 16

simulations were easily replicated with slightly better performance but the regulatory response showed much better results than what was shown in the paper. This is probably due to the fact that the input disturbance signal was not disclosed in the paper and it therefore had to be estimated. A discussion section presents an analysis of the results and a suggestion that this method of PID parameter tuning is best for systems which can handle overshoot, require a fast settling time and have built-in adaptive control. Finally a literature review is conducted and the current method is compared against a method provided by S. Skogestad [7] which gives tuning rules that provide predictable outputs for stable systems. When compared against the proposed method, the proposed method is more flexible but less robust and the S. Skogestad method is more robust but less flexible.

9 BIBLIOGRAPHY
[1] R. Padma Sree, M. N. Srinivas and M. Chidambaram, "A simple method of tuning PID controllers for stable and unstable FOPTD systems," Department of Chemical Engineering, Indian Institute of Technology, Madras, 204. [2] J. Bennett, A. Bhasin, J. Grant and W. Chung Lim, "PID Tuning Classical - ControlsWiki," 16 10 2007. [Online]. Available: https://controls.engin.umich.edu/wiki/index.php/PIDTuningClassical#ZieglerNichols_Method. [3] J. G. Ziegler and N. B. Nichols, "Optimum settings for automatic controllers," ASME Transactions, vol. 64, no. 759, 1942. [4] A. Abbas, "A new set of controller tuning relations," ISA Transactions, vol. 36, no. 183, 1997. [5] D. E. Rivera, M. Morari and S. Skogestad, "IMC-PID controller design," Industrial Engineering and Chemical Process, Design and Development, vol. 25, no. 252, 1986. [6] Q.-C. Wang, C. C. Hang and X.-P. Yang, "Single loop controller design via IMC principles," Automatica, vol. 37, no. 2041, 2001. [7] S. Skogestad, "Simple analytic rules for model reduction and PID controller tuning," Modeling, Identification and Control, vol. 24, no. 2, pp. 85-120, 2004. [8] A. Rad, MSE 490: Advanced PID Controllers - Lecture 4, Simon Fraser University: Surrey, 2014. [9] A. Rad, MSE 490: Advanced PID Controllers - Lecture 3, Surrey: Simon Fraser University, 2014. [10] G. H. Cohen and G. A. Coon, "Theoretical investigation of retarded control," Transactions of ASME, vol. 75, no. 827, 1953.

Page 17

APPENDIX A: MATLAB ERROR CALCULATION CODE


% Calculates the ISE, IAE and ITAE for a curve which function [ISE, IAE, ITAE] = CurveResults(Curve_t,Curve_y)

if(abs(Curve_y(end)) < 0.1) set_point = 0; %Set point for a disturbance rejection scenario else set_point = 1; %Set point for a servo response scenario end

% Calculate error error = Curve_y - set_point;

% IAE Calculation temp = cumtrapz(Curve_t,abs(error)); IAE = temp(end); clear temp % ISE Calculation temp = cumtrapz(Curve_t,error.^2); ISE = temp(end); clear temp; % ITAE Calculation temp = cumtrapz(Curve_t,Curve_t.*abs(error)); ITAE = temp(end); clear temp; end

Page 18

Você também pode gostar