You are on page 1of 12

MODEL REDUCTION BY EQUATING COEFFICINT METHOD

E Chandramohan, M. Chidambaram, Department of Chemical Engineering National Institute of Technology Tiruchirappalli 620 015, India. Corresponding Authors E-mail:chidam@nitt.edu

ABSTRACT In this work, a new model reduction technique has been proposed. The proposed technique is based on the equating coefficient (EC) method. Using this technique, stable third order processes have been reduced to SOPTD/FOPTD models with three tuning parameters. In any model reduction approach, it is very important that the approximated reduced model should resemble the original higher-order process. Extensive simulation studies have been carried out on different types of stable third order models. For small L/ ratio SOPTD models are desirable. In other situations FOPTD models are adequate. The proposed method is compared with the moments matching method and recently reported half rule method. Based on proposed method, the reduced model arrives straightaway by simply adding the time constants of the third order model. Keywords: Model reduction, PID controller, FOPTD, SOPTD 1. INTRODUCTION In general, transfer function model is needed for the design of controllers. In most cases, design of PID controllers is available for FOPTD/SOPTD models. Design of controllers with lower order models is more preferable than the higher order models. Thus in the design and analysis of higher order processes, it is often approximated to a lower order model like FOPTD models or SOPTD models. There are two approaches in the reduced order model controller design. In the first approach, the order of the plant is reduced and then a controller is designed for the reduced order plant. In the second approach, a controller is designed for the original higher order plant and then reduced. It is important that the approximated reduced order model should resemble the original higher-order process. Otherwise, the controller design with the reduced model will be inappropriate when applied for the original higher-order process. Methods of designing PI/PID controllers for stable FOPTD/SOPTD models are given by stability analysis (Ziegler-Nichols, 1942), IMC method (Scali and Rachid, 1998), Wang et al., (2001), Skogested(2003)),Synthesis method (Padma Sree and Chidambaram(2004)) and equating coefficient method(Padma Sree and Chidambaram(2004)). The main objective of this work is to design reduced order model for higher order processes. In this paper the equating coefficient (EC) method has been extended to in the designing a reduced order models (FOPTD/SOPTD) with three tuning parameters. Same tuning parameters use in the designing of a PID controller for third order plus time delay models and also procedure of designing reducing models(FOPTD/SOPTD). Researchers like Genesio & Milanese (1976), Al-Saggaf and Bettayeb (1990) have reviewed model reduction methods for reducing the higher order models to lower

order models. Model reduction methods such as Moments matching method (Lee, 1969) and half rule method (Skogested, 2002) are taking comparing with the proposed techniques. Stable higher order model with time delay is considered. The phase shift of the third order model 180,so with the PI controller is more than The controller considered is a PID with unknown parameters. In the closed loop transfer function (y/y r), the corresponding coefficients of s and s2 terms in the numerator are equated to 1 and 2 times of the denominator coefficients of the closed loop transfer function and k3 value is related to (i.e.k3 = this value varies depending on the L/ ratio). Because, in any method of controller design (like Ziegler- Nicholas method, IMC controller etc.) i value related to d. Thus by equating the coefficients, the unknown controller parameters are obtained from the known higher order model parameters. Considering these obtained controller parameters with the same 1 and 2, the unknown reduced FOPTD model parameters are obtained by equating the coefficients s and s2 of the numerator polynomial of the closed loop transfer function(y/y r), to the same 1 and 2 times with the corresponding coefficients of the denominator polynomial. Thus the unknown reduced order FOPTD model parameters are obtained. Figure 3.1 shows flow chart representation.

Figure3.1 Flow chart representation

2 THEORITICAL DEVELOPMENTS Equating coefficient method: Consider a third order model with time delay

( G ( s ) = k e Ls /( a s 3 + a s 2 + a s +1) ) and a PID controller with unknown parameters. p p 3 2 1 As the time delay term e-ls term in the numerator will only make a shift in the corresponding time axis it need not be considered in the analysis. However, the time delay term in the denominator cannot be neglected because it affects the stability of the system so it is approximated by using second order Pades approximation (Lam James, 1991). The closed loop transfer function relating the output variable(y), to the set point (yr) is given by

k 1s + k 2 + k 3 s 2 0.083L2 s 2 + 0.5L s + 1 y = y r s a 3 s 3 + a 2 s 2 + a 1s + 1 0.083L2 s 2 + 0.5Ls + 1 + k 1s + k 2 + k 3 s 2 0.083L2 s 2 0.5Ls + 1

)(

)(

) (

)(

(1)

Where k1 =k c k p ;

k 2 = k1 / i ; k 3 = k1 d

The objective of the control system is such that the closed loop response(y) tracks the set point (yr). The coefficients of s, s2 of the numerator polynomial can be equated with the corresponding coefficients of the denominator polynomial. With PID controller, the closed loop response shows some overshoot for the servo response, the value of y/ yr can be allowed more than 1. At s=0, y is automatically equal to yr, because of presence of integral action. The objective of the controller is make (y) equals to the (yr).Therefore, the corresponding coefficient of s term in the numerator is equated to 1 times to that of the denominator of the closed loop transfer function. The coefficients of s2 of the numerator are equated to 2 times of the denominator, the following equations are obtained. Here the value is related to i.e. k 3 = k 2 .
0.0 8 3 2 L2 + 0.5k 1 L + k 3 = 2 0.5L + a 1 + 0.0 8 3 2 L2 0.5k 1L + k 3 k k (2) 0.5k L + k = ( 1 0.5k L + k ) (3) Here and 2 are the tuning parameters. The choice of
2 1 1 2 1

the and 2 are based on many simulation studies. The values of and 2 around 0.52 and 1 0.88 give the best results. The above equations (3.3) & (3.4) become

From the above equations the unknown PID controller parameters k1 ,k2 and k3 are obtained. Hence kc, i and d also obtained. These are the controller parameters of the original higher order model. Consider the same known controller parameters (those are obtained from the above equations) and the FOPTD model (unknown parameters). The closed loop transfer function relating the output variable(y), to the set point (yr) is given by
( k 1s + k 2 + ks 2 )(0.083 L2 s 2 + 0.5Ls + 1) y = y r s( s + 1)( 0.083 L2 s 2 + 0.5Ls + 1) + (k 1s + k 2 + k 3s 2 )(0.083 L2 s 2 0.5Ls + 1) ( 6)

The corresponding coefficients of s in the numerator are equated times that of denominator and the coefficients of s2 the numerator are equated to times of the denominator of the closed loop transfer function. The following set of linear algebraic equations is obtained.
0.083k 2 L2 + 0.5k 1L + k 3 = 2 (0.5L + + 0.083k 2 L2 0.5k 1L + k 3 ) (7)

0.5k 2 L + k1 = 1 (1 0.5k 2 L + k1 )

(8)

Taking same tuning parameters of and 2 . The above equations become 1

From the above equations FOPTD process parameters are obtained. Design a controller on the reduced order model i.e. FOPTD model and implement that reduced order model controller parameters on original process. The servo and regulatory responses of original process controller parameters and reduced model controller parameters on original process are carried out. Consider the same known controller parameters (those are obtained from the (5) & (6) equations) 2 and the SOPTD ( K p e Ls / (s +1) ) model (unknown parameters). The closed loop transfer function relating the output variable(y), to the set point (yr) is given by
(k 1s + k 2 + ks 2 )(L2 s 2 / 12 + Ls / 2 + 1) y = y r s(s + 1) 2 (L2 s 2 / 12 + Ls / 2 + 1) + (k 1s + k 2 + k 3s 2 )(L2 s 2 / 12 Ls / 2 + 1)

(11)

The corresponding coefficients of s in the numerator are equated times that of denominator and the coefficients of s 2 of the numerator are equated to times of the denominator of the closed loop transfer function. The following set of linear algebraic equations is obtained.
0.083k 2 L2 + 0.5k 1L + k 3 = 2 (0.5L + 2 + 0.083k 2 L2 0.5k 1L + k 3 ) (12)

0.5k 2 L + k1 = 1 (1 0.5k 2 L + k1 )

(1 )3

Taking same tuning parameters of and 2 . The above equations become 1

From the above equations SOPTD process parameters are obtained. Design a controller on the reduced order model i.e. SOPTD model and implement that reduced order model controller parameters on original process. The servo and regulatory responses of original process controller parameters and reduced model controller parameters on original process are carried out.

3 SIMULATION STUDIES
In the design of the PID controller for the higher order models the tuning parameters 1 ,2 are around the range 0.52, 0.88, in the reducing the higher order models the same tuning parameters 1 ,2 and value depending on the L / ratio.Example1: The higher order process considered is Gp(s) = 4e s/ (3s+1) (2s+1) (s+1). From eq.(3.5), (3.6) the values of k1, k2 and =5 k2are obtained as 2.44, 0.623and 3.117 respectively. With this k1, k2 and k3 values, using eqs, (3.10) and (3.11) the corresponding reduced FOPTD model is obtained as Gm(s) = 4 e s/(6s+1). For this reduced order model, using the IMC settings (2003), the PID parameters are kc =0.541, i =6.5 and d = 0.4615.Fig.3.2 shows open loop responses of the original higher order model and for the reduced model. Fig.3.3, & Fig3.4 shows the corresponding servo and regulatory responses of two different controllers on original higher order process. The rise time and peak

time of both responses almost same but reduced controller settings on original model response somewhat sluggish because of less derivative action. Table 1 shows IAE, ISE values of closed loop servo responses Here the reduced model IAE, ISE values are close to original model IAE, ISE values. Example-2: The higher order process considered is Gp(s) = 4e 3s/(4s+1)(3s+1)(2s+1). From eq. (3.5), (3.6) and =k2, the values of k1, k2 and are 1.52, 0.2467 and 3.4545 respectively. With this k1, k2 and k3 values, using eqs, (3.10) and (3.11) the corresponding reduced FOPTD model is obtained as Gm(s)= 4 e 3s/(9s+1) For this reduced order model, using the IMC settings (2003), the PID parameters are kc =0.4375, i =9.5 and d = 1.285. Fig.3.11 & Fig.3.12 shows the corresponding servo and regulatory responses of the original higher order model and for the reduced order model. The rise time and peak time of both responses almost same but oscillations are more in second case because the product of k d is less same in the load disturbance also.
c

Table 1 shows IAE, ISE values of closed loop servo responses Here the reduced model controller on original model IAE, ISE values are close to original model controller IAE, ISE values. Example-3: The higher order process considered is Gp(s) = 4e 0.1s/(3s+1)(2s+1)(s+1). From eqs. (5.5) and (5.6), the values of k1, k2 and k3 =12 k2 are 6.9776, 0.4541 and 5.4492 respectively. With this k1, k2 and k3 values, using eqs, (5.10) and (5.11) the corresponding reduced SOPTD model is obtained as Gm(s) = 4 e 0.1s/ (3s+1)2 For this reduced order model, using the IMC settings (2003) of the PID parameters are kc =1.5, i =6 and d = 1.5. Corresponding reduced FOPTD model is obtained as Gm(s) = 4 e 0.1s/(6s+1). For this reduced order model, using the IMC settings (2003), the PID parameters are kc =0.5, i =6.05 and d = 0.05. Fig.5.1 shows open loop responses of the original higher order model and for the reduced models. SOPTD model open loop response very closely.Fig5.2 & Fig.5.3 shows the corresponding servo and regulatory responses of the original higher order model and for the reduced order model. Rise time and settling time is same in original and SOPTD but more in case of FOPTD. Table 2 shows IAE, ISE values of closed loop servo responses Here the reduced SOPTD model IAE, ISE values are more close to original model IAE, ISE values compared to reduced FOPTD model IAE,ISE and. Example-4: The higher order process considered is Gp(s) = 4e 0.1s/(4s+1)(3s+1)(2s+1). From eqs. (5.5) and (5.6), the values of k1, k2 and k3 =15 k2 are 6.8782, 0.581 and 8.715 respectively. With this k1, k2 and k3 values, using eqs, (5.10) and (5.11) the corresponding reduced SOPTD model is obtained as Gm(s) = 4 e 0.1s/ (4.5s+1)2 For this reduced order model, using the IMC settings (2003) of the PID parameters are kc =2.25, i =9 and d = 2.25.Corresponding reduced FOPTD model is obtained as Gm(s) = 4 e 0.1s/(9s+1). For this reduced order model, using the IMC settings (2003), the PID parameters are kc =0.5, i =9.05 and d = 0.05.Fig.5.4& Fig. 5.5 shows the corresponding servo and regulatory responses of the original higher order model and for the reduced order model. Rise time and settling time is same in original and SOPTD but more in case of FOPTD. Table 2 shows IAE, ISE values of closed loop servo responses Here the reduced SOPTD model IAE, ISE values are more close to original model IAE, ISE values compared to reduced FOPTD model IAE,ISE values

1.6 Original model Reduced model 1.4

1.2

Closed loop response,Y

Figure 2The closed loop servo responses for example-1.Solid line- controller design of higher order process: dashed line controller is designed
0.8 0.6 0.4

Figure-3. The closed loop regulatory responses for example-1. Legend same as in Fig.2

on FOPTD and applies on higher order model.


0.2 0 0 20 40 60 80 100 120 140 160 180 200

Time(sec)

2 data1 data2

1.5

Closedloop response,Y

0.5

-0.5

20

40

60

80

100

120

140

160

180

200

Time(sec)

2 .5 o in l rig a re u d o e d ce rd r 2

Closed loop response,Y

1 .5

0 .5

-0 .5

-1

2 0

4 0

6 0

8 0

10 0

10 2

10 4

10 6

10 8

20 0

Time(sec)

Figure-4. TheFigure-5. The closed loop regulatory closed loop servo responses for example-2.for example-2. Legend same responses Legend same as in Fig.2 in Fig.2 as

1.6 O riginal R educed m odel

1.4

Closed loop response,Y

1.2

0.8

0.6

0.4

0.2

0 0

20

40

60

80

100

120

140

160

180

200

Time(sec)

1 .4 o r ig i n a l F O P TD S O P TD

1 .2

Closedloop response,Y

0 .8

0 .6

0 .4

0 .2

20

40

60

80

100

120

140

160

180

200

T im e(sec )

Closedloop response,Y

Figure-6The closed loop servo responses for example6 .1re ,0 g 1 3.Solid.6 line- controller design of higher order original process: S PD OT 1 dashed.4 line controller is designed on FOPTD and F PD OT applies.2 on higher order model: dotted line- controller is 1 designed on SOPTD and applies on higher order model. 1
0 .8

Figure-7. The closed loop regulatory responses for example-3. Legend same as in Fig.6

0 .6 0 .4

0 .2 0

-0 .2

2 0

4 0

6 0

8 0

10 0

10 2

10 4

10 6

10 8

20 0

Time(sec)

Table-1: sum of the error values of servo response


Controller design of higher order model ISE(IAE) Controller design of FOPTD on higher order model ISE(IAE)

Example-1 5.1113 (7.40) Example-2 10.5508 (15.9289) Table-2: sum of the error values of servo response
Controller design of

5.0307 (8.2539) 9.76(16.6141)


Controller design of

Controller design of

higher order model ISE(IAE)

SOPTD on higher order model ISE(IAE)

FOPTD on higher order model ISE(IAE)

Example-3 Example-4

2.6926(4.1374) 3.7063(5.8032)

2.7558(4.1588) 3.3107(5.3689)

3.7163(5.6006) 5.2202(8.4717)

4 COMPARISON STUDIES:
Example-5: The higher order process considered is Gp(s) = 4e s/ (3s+1)(2s+1)(s+1). From eq. (4.5), (4.6) the values of k1, k2 and =5 k2are 2.44, 0.623and 3.117 respectively. With this k1, k2 and k3 values, using eqs, (4.10) and (4.11) the corresponding reduced FOPTD model is obtained as Gm(s)= 4 e s/(6s+1) Reduced model obtain from the Half rule method as. G m1(s)= 4 e 3s/ (4s+1).Reduced model obtain from the Method of moments as Gm2(s)= 4 e 3.258s/(3.742s+1) Fig.4.2 shows the corresponding servo responses of the original higher order model and for the reduced order models. The rise time and peak time of both responses by using EC method almost same but in case of reduced model by using half rule and method of moments the response become over damped because of the less kc value and reduced controller settings by using EC method on original model response somewhat sluggish because of less derivative action. Table 3 shows IAE, ISE values of closed loop servo responses .Here the reduced model IAE, ISE values by using EC method are very close to original model IAE, ISE values compared with half rule and moments matching method IAE, ISE values. Example-6: The higher order process considered is Gp(s) = 4e 3s/(4s+1)(3s+1)(2s+1). From eq. (3.5), (3.6) and =k2, the values of k1, k2 and are 1.52, 0.2467 and 3.4545 respectively. With this k1, k2 and k3 values, using eqs, (3.10) and (3.11) the corresponding reduced FOPTD model is obtained as Gm(s)= 4 e 3s/(9s+1) For this reduced order model, using the IMC settings (2003), the PID parameters are kc =0.4375, i =9.5 and d = 1.285. Reduced model obtain from the Half rule method as Gm1(s)= 4 e 6.5s/(5.5s+1). The reduced model obtain from the Method of moments as Gm2(s) = 4 e 6.615s/ (5.385s+1) .Fig.4.1 shows the corresponding servo responses of the original higher order model and for the reduced order models. . The rise time and peak time of both responses by using EC method almost same but in case of reduced model by using half rule and method of moments the response become over damped because of the less kc value. Table 3 shows IAE, ISE values of closed loop servo responses. Here the reduced model IAE, ISE

Closed loop response,Y

values by using EC method are very close to original model IAE, ISE values compared with half rule and moments matching method IAE, ISE values.

1 .6 orig l ina EC M ha lf m m ts o en

1 .4

1 .2

0 .8

0 .6

0 .4

0 .2

20

4 0

6 0

80

1 00

12 0

14 0

16 0

1 80

20 0

Time(sec)

1.6 Original EC method Half rule Method of moments

1.4

1.2

Closed loop response,Y

0.8

0.6

0.4

0.2

20

40

60

80

100

120

140

160

180

200

Time(sec)

Figure-8The closed loop servo responses for example6.Solid line- controller design of higher order process: dashed line controller is designed on FOPTD and applies on higher order model (EC method): Dash-dot linecontroller is designed on FOPTD and applies on higher order model (Half rule): Dot line- controller is designed on FOPTD and applies on higher order model (Method of moments).

Figure-9. The closed loop servo responses for example-5. Legend same as in Fig.8

Table-3: sum of the error values of servo response Controller design of higher order model ISE(IAE) Controller design of FOPTD(EC method) on higher order model ISE(IAE) Controller design of FOPTD(Half rule) on higher order model ISE(IAE) Controller design of FOPTD(Method of moments) on higher order model ISE(IAE)

Example-5 Example-6

5.113(7.40) 10.55(15.93)

5.0307(8.2539) 6.9823(10.0861) 7.3500(10.8411) 9.76(16.6141) 10.724(14.061) 10.834(14.295)

5 CONCLUSIONS: Higher order models are reducing to SOPTD models more appropriate than reducing to FOPTD models with the and 2 tuning values around 0.52 and 0.88, value varies depending on 1 the L / ratio. (When higher order processes having small delay with large time constants). The procedure of the proposed method is equating the numerator and denominator coefficients of the s and s2 terms in the designing a PID controller for third order model and in the reducing model also. The equating coefficients s term in the designing a PID controller and in the reducing model both are same. The equating coefficients s2 terms only the difference is a1 (sum of the time constants) came in the designing a PID controller and in the reducing model and time constant 2 came. To reduce the complexity in the derivation, the reduced model simply written by half of the adding the all time constants of the denominator and retain the numerator as such for examples are considered

6 REFERENCES

1 Al-saggaf .U .M and Bettayeb .M; Optimization in model reduction. The Arabian

journal of Science and Engineering, Volume 15, October 1990, Number 4B 705-719
2 Astrom, K. J., T. Hagglund, C.C Hang and W. K. Ho (1993) Automatic tuning and

adaption for PID controller- Control Eng. Prac., 1, 699-714.


3 Bequette B. W 2003.Process Control -Modeling, Design, and Simulation

Prentice-Hall, Inc.,
4 Chidambaram. M and Padma Sree. R. (2003); A simple method of designing PI/PID

controllers for integrator/dead time processes, Comp. & Chem. Eng, Vol-27, 211- 215.
5 Coughanowr. D. R.(1991). Process Systems Analysis and Control, 2nd edition,

McGraw-Hill, Inc.
6 Douglas J. M (1972)Process Dynamics and Control, Volume-1, Prentice- Hall,

Inc.,
7 Genesio .R and. Milanese M, (1976)A Note on the Derivation and Use of Reduced

Order Models, Transactions on Automatic Control, AC-21, p.118.


8 Gibilaro L.G. and Lees F.P(1969); The reduction of transfer function models to simple

models using the method of moments. Chemical Engineering Science, , vol.24, pp, 85-93.
9 Kaya. I(2004); IMC based automatic tuning method for PID controllers in a Smith

prediction configuration, Computers and Chemical Engg. 28 281-290..


10 Lam. J1991; Balanced realization of Pades Approximants of e sT IEEE Transactions

on Automatic Control, Vol. 36, no. 9, September.


11 Padma Sree .R and Chidambaram .M (2006).Control of Unstable systems,

Narosa Publishing House Pvt. Ltd.


12 Padma Sree, R., Srinivas M. N. and Chidambaram .M,; A Simple method of tuning

PID controllers for stable and unstable FOPTD systems, Comp. & Chem. Engg., 28. 2201-2218.
13 Skogested .S(2003); Simple analytic rules for model reduction and PID controller tuning.

journal of process control 13 291-309


14 Ziegler J.G., Nichols N.B(1942)., Optimum settings for automatic controllers, Trans.

A.S.M.E. 64 759-768.

You might also like