Ish tranacsone 56 (201) 30 Contents lists available at ScienceDirect ISA Transactions journal homepage: www.elsevier.comilocate/isatrans Two degree of freedom control scheme for unstable processes Dens with small time delay Moina Ajmeri*, Ahmad Ali Indian Inte fTeclgy Pa, nin ARTICLE INFO ABSTRACT Ice nat Ti ths work + recently reported two degre of freedom control rrucare namely the Parallel Conta Received 20 June 2018 Structure (PCS s modified to achieve improved contol performance for unstable processes, New tuning Received in revised form, res are proposed using a direct synthesis approach for unstable fist order pls time delay (UFOPTD}. Meee: 2g Lnstabl Second onder pls time delay (USOPTD) and integrating and unstable plus tims delay (IUPTD} ‘eae one ST ouary 013 process models. The proposed method has two tuning parameters whose suitable values are recom- Irina a ecoamee lot ‘ended after studying their effect on the system performance and robustness Tiss an advange ofthe pean by Dr Als Bad present work over th other reported contol schemes were the authors provide suitable eanges of the ER fuming parameters values Simulation examples show that the present scheme results in improved ewer ‘nomial and perturbee sponses as compared othe recently reported methods. Also its observed that Sesable proces Satstatory performance is achieved when the proposed tuning rules te applied on an isothermal uel Contol Suc hema! reactor which exits multiple steady state solutions. 2 © 2015 ISA Published by lever Le Ali reserve. 1. Introduction A.comprehensive summary of PY/PID tuning methods reported inthe literature fora class of processes is given in [1] It can be observed. from [1] that the number of tuning rules is much more for stable processes as compared to unstable process models. Several works have been, therefore, reported in the literature for designing PID and its variants for open loop unstable processes [2-7], A PIPID controller ‘within the framework of a unity feedback control structure results in closed loop step response with excessive percentage overshoot and large setting time for unstable processes. To overcome this problem, two degree of freedom control schemes with double feedback loops have been reported in the literature [5~10), The inner loop controller was designed to stabilize the unstable process model whereas the desired system performance for the augmented process was achieved by tuning the controller inthe outer loop. The inner loop PD and the ‘outer loop PI contrllers were tuned to meet gain and phase margin specifications for stable and unstable frst order plus time delay ‘process models in [8]. Nie etal. [9] have designed a leagjlag compensacor in the inner loop using stability margin (gain and phase margins) specifications and a Pl controller in the outer loop using the method reported in [11], Te outer loop PID controller was tuned using an Internal Model Control (IMC) based technique for stable. integrating, inverse response and unstable processes in [10], The inner loop controller in the above cited work was designed using either arelay feedback approach or a Ziegles-Nichols method. Tuning rules for both integer as well a5 fractional PID have been reported in [12] by minimizing the IAF (integrated absolute error) performance criterion for integrating plus time delay and first order unstable processes. The above discussed two degree of freedom control schemes have ‘considered processes with only one unstable pole The two degree of freedom control structure proposed by Lit et al. [13] is shown in Fig. 1, where G is use for stabilizing the delay-free part of the process model (Gyo). The controllers C and F are employed for tracking the Set-point and rejecting the load disturbance, "espectively, Tuning rules based on Hs optimal performance (which is equivalent to integral squared error performance specification) were reported by the authors forthe following unstable fist order plus time delay (UFOPTD), integrating and unstable plus time delay (IUPTD), and unstable second order plus time delay (USOPTD) process models: ‘6019-0578/6 2015 1A, Published by Elsevier 1g Al rights reserved,LM Aimer A.A / 1 Isooone 5 (205) 208-126 309 ror: Gms = w wuero Ga Re, @ sort Gus)“, % Tess Subsequently, Shamsuzzoba and Lee [14) have proposed new tuning rules forthe disturbance estimator (F| shown in Fig. using an IMC based method for IUPTD and USOPTD process models. A PID in series with a lead-lag filter was designed using a direct synthesis approach in [15,15] for the above said process models, Furthermore, the authors in [14-16] report that their methods give improved closed. Toop performance as compared to Liu et al's [13] strategy. The main drawback in the above cited direct synthesisIMC based works is that they provide a starting value of the tuning parameter which needs to be fine tuned to achieve the desited closed loop performance and robustness, Recently, Vanavil and Rao [17] have proposed PID cascaded with a higher order lead-lag compensator for first and second order unstable processes. Suitable values of the tuning parameter have also been recommended in the above said work. ‘A.wo degree of Ireedom control structure known asthe Parallel Control Structure (PCS) has been recently reported in (18). The general form ofthe PCS which is shown in Fig. 2 decouples set-point tracking response from the load disturbance rejection response. The authors have considered only stable and integrating processes in [18] to illustrate the working of the PCS. In the present work, inspired by Litt, et al’s [13] method, the PCS is modified by introducing an additional controller to stabilize the delay free part of the process model. Furthermore, controllers of the modified PCS are designed using the direct synthesis approach for UFOPTD, IUPTD and USOPTD process ‘models, The direct synthesis approach has been used because it enables the user to achieve a trade-off between the performance and robustness by adjusting a single design/tuning parameter. Process models having normalized time delay less than or equal to one have been considered in this manuscript. This is because P1/PID controller falls to give satisfactory performance for plants having normalized time delay greater than one. Smith predictor and its variants [19-22] have been reported in the literature for the control of above said processes. The main contributions of the present work can be summarized as fllows: (1) A modified PCS is proposed which results in improved closed loop performance for unstable processes. New tuning rules ate reported {for UFOFTD, [UPTD and USOPTD process models using the direct synthesis approach. It isto be noted that only few of the reported ‘works 13,17 give controller settings forthe above said class of process models. Furthermore poor load disturbance rejection has been reported in |13| and [17], However, the proposed scheme gives good naminal and robust performance for all the considered process models, (2) Itis observed from the literature survey that none ofthe reported works except [17] provides suitable values of the tuning parameters. Inthe present work, after studying the effect of tuning parameters on the system performance and robustness, suitable values of the ‘tuning parameters are recommended for a class of unstable processes, Om « Fie Tw dept of reedom conto srute by Ut [12 Ge Ga0 1M Aimer. Ah I rsotone 5 (2015) 208-126 ‘The paper is organized as follows: modified Parallel Control Structure is discussed in Section 2 whereas direct synthesis based design of ‘the modified PCS is presented in Section 3. The robust stability and robust performance conditions in the presence of process multiplicative uncertainty are studied in Sesion 4, Furthermore, suitable values ofthe tuning parameters are recommended after studying ‘their effect on the system performance and robustness in this section. Simulation exaraples are given in Section and conclusions are drawn in Section 6, 2. The modified Parallel Control Structure The PCS shown in Fig. 2 incorporates both the nominal model of the plant (Gy) as well as the actual process (G,) which is to be controlled, It has two controllers namely Gy and Ge, whose control signals are u and t, respectively. Gy, takes care of set-point tracking response whereas the load disturbance is rejected by the controller Gz. In the present stuéy. the PCS is modified by adding a third controller 'G¢ to stabilize the delay free part of the process model. The modified PCS is shown in Fig. 2. Gye tepresents the delay fee part of process model Gy and the time delay is indicated by e~*. The closed loop response (y) of the modified PCS is given by 1 GaGe Gab r+ [_& Te GexGy) [TT EG + Cam" * [TFCaGe ‘where all the signals and systems are in Laplace domain. The symbols r,y, and d represent the set-point, controlled variable and the load disturbance atthe plant input, respectively. Under nominal conditions (CpG), ¥ can be expressed as follows: Can Ge + fetta lel! ® 4 “ Prom Eq, (5), it can be observed thatthe set-point tracking is taken care of by the controllers Gy, and G. while the load disturbances are rejected by the controller G,2. The modified PCS therefore decouples the regulatory (load disturbance rejection) response from servo (set- point tracking) response. 3. Controller design In the present work, the controllers of the modified PCS are designed using a direct synthesis method. In the direct synthesis approach, the controllers are designed to achieve the desired closed loop transfer functions for set-point changes or load disturbances [15.17 23,24] For specifying the desired closed loop transfer functions, the actual closed loop transfer functions are obtained under nominal conditions ‘of the modified PCS. The delay term in the denominator ofthe obtained transler function is approximated using Pade's or a povrer series approximation so that the order of denominator becomes equal to the number of controller parameters, The numerator of the desired ‘transfer function is set equal to the ntimerator of the transfer function obtained under nominal conditions, The desired closed loop time constants for the set point and load disturbance responses are assumed as 4 and r, respectively in the present paper. 34, Setpoint tracking controllers (G, and Gis) G_ =k, and Gy =Kyy (1+ 1/T,s) are considered for the UFOPTD process model whereas G- and Gy, are assumed as ke(1-+t,8) and 1/Tyys {or IUPTD and USOPTD process models. The following conditions must be satisfied so that the roots of the inner loop characteristic ‘equation (1-+GyoG, =0) lie in the left half of the s-plane: (a) UFOPTD: ke > 1/K (b) PTD: ktg> 1/K, and Ki > 0 (6) USOPTD : Kke> 1, Kketg> (Ty +Ty) and Ts > 0 ‘LM Aimer Alt Ison 5 (205) 208-126 am 3.11, UFOPTD process model Substituting Ge = ke, Gr =Kyi(1 +8). Gn = (K/TS—1) and Gn =K(e™**/(T5— 1) in (5). we get y (Ke? Ts) (1 +O/T 8) TO TE RR TS =I + RE=PTS DR TA @ Replacing e~** by (1-050 s)/(1+05505) in the denominator, we get y e414 Ts5)(1+0/29) 7 Ty TOPARR PF (Tal RR) OPARK a) + gk OPIK ps) =O) 2)|S > (a OPA = Ta RR Take Kes ® Let the desired transfer function be a Sha Tys(1410/2 Peres eee ® By comparing Eqs (7) and (6), the following controler settings ae obtained for the UFOPTD process model fea Mawnan enna ven Kym (arg 2 ah ® Hi 3:12. IUPTD process model Substituting G,=ke(1-+t48). Gey =(1/Tas). Gy (1-05), we get the following expression for y/r (e-® /STS—1)), Gye =(K/s(TS—1) and replacing e-* in the denominator by y a" 1 TT RP FURR DRT eT OST oo The desied transfer function is assumed as follows 6 ee ay eniee™ Gay By comparing (10) ané (11), we get eo t= (8) (2) Tha 313. USOPID process model “The following controller settings are obtained for the USOPTD process model using the approach discussed in the previous subsection: 4, 02 (pane) ay Ta=# 32. Load disturbance rejection controler (G2) “The closed joop transfer function between the output and load disturbance at the plant input (under nominal conditions) is obtained as follows: ad ([T¥GaGm) « eel) (ua Gat)=Kn (1+) GER) a5) va (tetoor (1) , Gath (1+tat+q) ($28) oo for the second order processes.By 1M Aimer. AA 1 rasan 5 (2015) 208-126 3.21, UFOPTD process model Substituting (1) and (15) in (14), we get y Keo /Ts—1) a ET 1) TERR FTES VTC Tas aE uo Replacing the exponential term in the denominator by Pade's approximation (as used in Section 3.11) and assuming a-=62, we get the following expression: y Tas(1+fs)e-"(1/Kya) 7 TTR TaN RR NT = BY DT Ta RR aT OAT om The desired transfer function is assumed as 1 (4 )e-Pks (2, Taliesin a a By equating the denominator coefficients ofthe above two transfer Functions, we ge the following tuning rules Kw 282908 46M eso BROT 30T OT /2)K 4, 2P18TH 6 OF a= Srey ° 207-02 an, po) 20 2 Psy ere oT OTR eo 322. 1UPED process model ‘yd for the IUPTD process model is obtained as follows eS ee ep 7 TARR PES DIT Tas Ta SOTO E) a b rt) i dS Teme) Fg (5) fet of onthe serve response for Gs (nominal) (0) Efe onthe cont sgl fr Gay (minaLM Aimer A/S Tnsooone 5 (205) 208-126 oy Assuming e=° = (1-0.508)/(1 4 0.50s) and c-=6)2, we get K(14ps)Tase~*/KKa TTaDHTRR a) (Ta TAB RR (TaTn0)2))9 + (Tat (TaO2)— al Rae ‘The desired transfer function is assumed 36 (9/4) gang = ((1 Bs) Tose" (Ky) £9 which results in the following controler stings for Ga 2) Wart ener 0) Tae sere PTT) Ta =4r+0/2 3 ao 423, USOPID process model On salar ines the following tuning formulas for Gq are obtained for USOFTD process models Ka x 2 (ieheoneannioye Tarren Mh reo where x= (—0-42T; +272) e4/TsT20+ 8 (0-4 602+ (Ty 4T,—27sT2/0) 40+ 0/2) >. i ly Tame) ig. 5, tect of. on the serve response fons (pebble onthe conta sgl fr Gs (perturbedae 1M Aimer. A Ah I rasan 5 (2015) 208-126 Remark 1. For second order processes, the phase lag imposed by the term (fs 1) is large and hence, the proposed controller settings fil to give satisfactory system robustness. tis reported in 4 an [15] that robust performance can be achieved for second order processes without any zero Dy replacing f with 0:18. Hence, 016 instead of fis used in the present work Fig. 7 (a fet of onthe eiturbance response for Gn namin (b) et of onthe etrbance respons for Gn pareAme A. Al SA Iron 56 2015) 308-326 as Remark 2, In the present paper, tuning rules for USOPTD processes are obtained by considering Ty > 0 and Ts» 0, It isto be noted that these controller settings are also applicable to second order processes with one stable and one unstable pole. Suppose the process to be controlled is as follows Ont RTD a ‘The proposed tuning rules for USOPTD processes can be applied tothe above process model by assuming K-= -K, Ty =T, and T= To 4, Robust stability analysis ‘The plant model which i used for designing a controller is only an approximation ofthe true dynamics ofthe actual system. Hence, tis necessary to analyze the robust stability ofthe closed loop system for uncertainties in the pracess and load disturbances, According to Karunagatan and Wenjian [18 the closed loop PCS is robustly stable if and only if (1-86 gan) Anda) <1 Vere (20.00) 26) where (6 > Go) Gq) = Gaara) 7) (1-s6> 0 CaGeNGao} an which represents the complementary sensitivity function under perfectly matched condition and AGa) ~ (Gio) a) (Gna is the bound on the process multiplicative uncertainty. Assuming (1-5 Gj) = Cain, Ea. (26) reduces to NCHA WGe} oe <1 Yeo 28)a6 1M Aimer AA 1 rasan 5 (2015) 208-126 Using Eqs. (1), (15) and (27), we get the following closed loop complementary sensitivity function for the UFOPTD pracess model Kkype-Tas-1Nas+1) Hi) = og ne ast as a: 0) = TNs yas Kaas Mas TE @ uncertainty exists i the process time delay, the tanng parameter + shouldbe selected such that ICG ) EF of ron the esturbance respons er Gn (peru ig. 12, Masimum senstiiescrrespending to = 1.1286 Gna IM Aimer. Ah I rsotone 5 (2015) 208-126 ‘observed for other UFOPTD pracess madels. Hence, 28 is recommended for UFOPTD processes, On similar lines, 2=18@ and A=@ are recommended for IUPTD and USOPTD process models, respectively, 42, Selection of = Decreasing improves the load disturbance ejection performance but degrades the closed loop's robust stability in the presence of the process uncertainty. On the contrary, as + increases the robust stability improves and the disturbance rejection response deteriorates. The tuning parameter 7 should satisfy Eqs (28) and (33) to achieve robust stability and robust performance. Based on the extensive simulation studies, suitable Values of r are recommended so that a small value of maximum sensitivity is achieved, Maximum sensitivity (M,) is defined as the inverse ofthe shortest distance from the Nyquist curve ofthe loop transfer function to the critical point (-1, 0) It isto be noted that a small value of M, implies that the system is more robust to the parameter perturbations. 42.1. Processes with only one time constant (Gy and Gyz) ‘The values of the maximum sensitivity obtained by varying + are shown in Fig. 6 for UFOPTD process models. Its observed from the figure that as 7 increases. M, decreases and therefore the system becomes more robust to process uncertainties for 8/T'<0S. To demonstrate the effect of + on the system performance and robustness, an UFOPTD process model with transfer function Gyx =e~® /(5—1) is considered. The nominal and perturbed (for an uncertainty of +20% in 0) responses for a unit step load disturbance ate shown in Fig. 7 ig. 7a) shows that r= 110 results in better closed loop response as compared tothe other two values afr. However, it ‘is observed from Fig, 7b) that the robustness corresponding to r= 110s not satisfactory. This i also evident from Fig. 6, M, corresponding to f= 110 is large. Good nominal performance and satisfactory robustness are achieved for r= 1.36, Furthermore, itis to be noted that ‘£130 results in satisfactory closed loop performance and robustness for process models with 8/T-< 0:5 Als, the variation in M, is very small for ¢> 130 for the above said range of 8JT. Therefore, i is recommended to use +=138 for O/T = 05. From Fig. 6, itis observed that r=1.70 results in minimum M, for 8/T=0.7 and 0JT=0°. Also, the value of M, corresponding to r=1.70 is small for 6/T=10, Hence, 7=1.78 is recommended for 8(T >. ‘The maximum sensitivity plots for UPTD process models are shown in Fig. 8. It can be observed that r=180 gives minimum maximum, sensitivity for process models with 8/7 > 05. Furthermore, ¢=1.80 results in satisfactory nominal and robust performances for process ‘models having 0/P'= 05. Therefore, r ~180 is recommended for IUPTD process models. 422. Processes with two time constants (Gua and Gna) ‘The USOPTD process models (Gps) have two time constants Ty and Ts. Assuming? — TyTs/(Ty (Ts), the maximum sensitivity pots are shown in Fig. 9. The effect of T; and T, on M; is explored and itis observed that different combinations of T; and T; that result in the same {)T give almost the same M, for a given r, The system ouputs for a unit step load disturbance for Gyo = (2e-°% (3s 1)(s—1) are shown. in Fig. 10. The perturbed responses are obtained by assuming +-20% change in 8, It can be observed from Fig, 10{a) thatthe disturbance rejection performance degrades as ris increased from 1.230 to 1430. However, a large value of + implies improved system robustness as is Le eT CT Macs es Gaidelines fo selecting «Ame A. Al SA Irnectins 562015) 308-326 a9 evident from Fig, 10) It is toe noted that z= 1438 results in poor nominal performance whereas 21330 yilds satisfactory robustness and good disturbance rejection response, Similar trend is observed for ather USOPTD process models with 8/7 = 05, Hence, =1330 is Fecommended for 6/T = 05. On similar lines, 1530 is selected for USOPTD process models having 05 ~< 0/7 = 08. ‘The controller parameters for Gps =e~8*/((6—1\05s+1)} ate obtained by substituting T; =05, 6-05 and K. 4. (24). The nomial and pesturbed (30% change in 8) responses for an inverse unit step load disturbance ate shown in Fg. 1. The system robustness corresponding to *=1030 isnot satisfactory, wheseas ¢=1130 result in good nominal and tobust performance forthe above considered process model. Fig. 12 shows the maximum sensitivity plots coresponding o #=1130 fora range of T/T, values. tis observed from Fig 12 that M, varies within the rage 2-5 which is acceptable for unstable processes. Hence, :=1130 is recommended for 02<|fe/Ty)21 a i Frequency eee) reposed complementary sensivty function ané that af 1/(e~#1* =1) fer Example 1. ° . Teng mods Se polar wading Ditubaace econ Vijayan and ands [1 2 130 035 2 570 as ia ae Pada nd Vis [12 aa at ae aise Sas to4 037 339 Vani and a 17 6 a a0 on 18 oa ea 426 Visvan ane Panes |] nas toe no 338 72 38 a 07320 1M Aimer Ah 1 rnsotone 5 (2015) 208-126 Working on similar lines, itis observed that +=1.538 gives satisfactory performance and robustness for process models having. 1.<|To/Ts| $ 10. Values of the maximum sensitivity corresponding to t=1,536 ate shown in Fig. 13. The recommended values of¢ for the considered class of unstable processes are summarized in Table 1 5. Simulation study ‘The pure derivative terms are implemented by cascading them with a fist order low-pass filter with time constant equal to 0.01 times ‘the derivative time constant. The integral of absolute etror (JAE) integral of squared error (ISE, setting time () and total variation (TV) are calculated to compare closed loop performances of various tuning methods. Mathematically, ISE and IAF in the controlled variable are a | Tine se) c ig 1, (Step responses fo Example 2 namin fb} Cote signals frExame 2 (nominal) (Sep rspanss fr Example 2 (perturbed.Aime A. Al /ISAIrnectine 562015) 308-325 a sivenby mae [ewe oy tse [eee 2) In the above equations, e(t is the difference between the set-point and controlled variable. A small value of JAEJISE implies fast set- point tracking and load disturbance rejection Setling ime is the lime requited by the step response to settle within 2% ofits steady state ‘value, The total variation (TV) of manipulated variable isthe sum of all control moves, both up and down, if ult)is discretized as a sequence [us ua usaf then TVs given by 1 Sins. 65 where TV is a measure of the smoothness of control signal and it is desired that TV be as small as possible (25} Example 1. Vijayan and Panda (10) have proposed a proportional controller with gain equal to 1.268 in the inner loop, Grp =0.3533(1-+ (1/1.50465)+.0.5166s) in the outer loop and a set-point filter with time constant equal to 04 for the UFOPTD process model Goy(s)=e79%/(5—1), A PL controller in series with a second order lead lag filter, Gyn = 15741(1-+(1/10.1449s)) ((140.25s+ (0.02085%)/(1-} 0.0851s0.00827s")) was proposed for the above said process model in [17]. The controllers reported in (12) for set-point tracking and load disturbance rejection are Gy ~2.1011(1-+(1/1.9324s)) (0.27728 1)/(002772s+ 1) and Ge, = 2.09381 + (1/1391 19) (030755 + 1)/(0.03075s+1)), respectively. For a fair comparison, the above controllers have been used in the two degree of freedom structure shown in Fig. 2. The proposed controller settings are ke=4, Ky, =05, Ty =025, Kyo = 17887, Ty = 48895, 025, and p 0.0984. Te closed loop performances achieved by the above said methods are compared by applying a unit step change in the set-point at time (=0 and an inverse unit step load disturbance at the plant input at time C=15. The resulting system outputs and control signals are shown in Fig. l(a) and (b). The performance measures are given in Table 2. fig. 14(c) shows the system outputs obtained by assuming a perturbation of +20% in the process time delay. Paula and Visioli’s (12) method fails to stabilize the system and hence is not included in, Fig, 14(c) Its observed from Fig, 14(a) and Table 2 that the strategy reported in |17) results ina larger peak value of the system output and longer settling time as compared to the proposed scheme for bath set-point and load disturbance responses. Vijayan and Panda's (10] method gives better servo performance as compared to the present method, However, for the load disturbance rejection, peak of the system output, JAE and ISE corresponding to the proposed work are small compared to the approach given in [10] It is to be noted that Toad disturbance rejection is more important than set-point tracking in many control applications [14] In the present study, suitable values ofthe tuning parameters are recommended which is an advantage over the method reported in [10]-In Vijayan and Panda's (10) method, the user has to adjust the tuning parameter until t gives less ITAE value. Fig. 14(€) shows the magnitudes of complementary sensitivity function and 1/(e~®"*—1) which corresponds to an uncertainty of 20% in the process time delay. It can be observed that the proposed strategy satisfies the robust stability condition given by Eq, (20), Example 2. The controllers reported in |14) for the [UPTD process model, Gna = (€~22/s(6—1)) ate Ge= 1425, C= (2 44 1)/(O6e+ 8 and F =3,0241(1 + 1,0585-+(1/1.7941s)}(0.10s +1)/(0.0087s + 1), Nie et al. [9| have proposed Cis) — 0.714.725 + 1)(0.00465 + 1) in the inner loop and C.(5) = (0.3025 +1,3236/5) in the outer loop for the considered process model, The controller and the set-point weighting. parameter reporied in [15] are 2.584(1 + 1.274s +(1/2.44s)}(0.10s-+1)/(0.0044s + 1)) and 0.536, respectively. The present method gives ke =27.4348, 4 =0.3402, Ty =0.0467, Ky) = 4.8733, Ty =15400, Ty =08189, a=0.1, and f=00051 Fig. 15(a) shows closed loop system outputs for a unit step change inthe set-point input at = and a negative step load disturbance of magnitude 2 applied at ¢= 15 at the process input. The corresponding control signals are shown in Fig, 15(b)- Nie et l’s [9] method is not included in Fig. 15(b) as it results ina large initial change in the control signal for the assumed change in the set-point. Perturbations of 20% and -20% are introduced in the process time delay and time constant and the resulting system outputs are shown in Fig 15(c). Its to be noted that Nie et al's [9) method gives unstable response for the assumed perturbations in the process parameters. Various Teng mods Sept wading Disubance recon Neal (3) zor oa oat 2534 2031 ais bat 1s8 Shansuzoba and Lee (4) aor vat oe ‘460 415 120 os 77 oes chidambarsr 9) a7 18 os 1 oa te ae soe Proposed 304 132 os om as os on Nee (9) - - - * * . . Shameuezoa and ee [141 sor ia ast 778 ass a0 age anand Chdambarac [15] 52 ur oa teat a 130 beta 1M Aimer Aah I rsotone 5 (2015) 208-126 performance measires are given in Table 3. [ts observed from Fig. 15{a) and (c) and Table 3 that the proposed scheme outperforms other strategies under both the nominal and perturbed conditions. Example 3. The authors in [15] ané [17] have proposed Gp =2.972(1 + 1.6145 +(1/1.85) (1155+ 1/(0.004184-1)) and Gy 2.7255 (15 1.73695 ~ 1/1.98375) (0.00255 4+-0.15s= 12.0022 01032551) forthe USOPTD process model. Gyp =2e~2> /(35— 1)~ 1. Also, set point weighing of 0536 and 03 have ben used in the above works to remove the peak overshoot. The coolers designed by Shamswzzoha and Lee [14] ate Ge=35, C=(1.5s°+5+05)/(051s+1)" and =3.5671(1+ 1.33645 (1/1.4913)} ((0.15s+1)/(0.0058s-+1)). The present tmethod gives k= 66.1067, t4= 02569, Ty =0018, Ky: =40076, Ty= 1.9523, Tay= 1259, @=0.15 and f= 0.0044 ‘The system outputs and contol signals obtained fora unit step change in the set-point at 0 and an inverse step load disturbance of magnitude 2 at time t= 15 are shown in Fig, 16(a) and (b). Ie is observed from Fig. 16(a) and Table 4 that che proposed strategy results inthe least ¢, for set-point tracking. Shamsuzzoha and Lee's [14] method gives minimum settling time for rejecting the load disturbances. a Tara Tine) Cont vse tas e ost ff i all Tine (es) ig. 16 (8) Step responce for Earle 3 (nominal (b} Canto sgal fr Example» namin). (Step responce fr sample perbe