Chapter 1 9 | Non-Linear Control Systems ehave already introduced linear control systems and studied them from their pole-zero configuration, «hich is an extremely powerful tool both in frequency and time-domain analysis and synthesis. Unfor- tunately, systems that exist in nature are non-linear. Hence, the superposition principle which is well «alid in linear systems is totally invalid in non, systems, There are control systems which do not taveany linearized version and itis very difficult to linearize these control system. In some eases, it has seen found that the non-linearities such as dead zone and hysteresis and combination of these are unde~ sable and unavoidable. The aim of this chapter is to discuss the methods to study non-linear systems aswell as their stability. 19.1 The State Space Approach Figure 19.1 shows the block diagram representa dynamics is represented by the state vector x. In controlled output vector. Controller u Plant — as ystem tion of a class of non-linear control systems. The plant Fig. 19.1, ris the reference, wis plant input and ¢ is the Fig. 19.1 Non-linear controls The mathematical models to represent the plant and controller are given below. Play x= fio) (19.1) a ~=Ce (19.2) Cntr otlep uae” (19.3) Scanned with CamScanner798 | Control Systems in peneral non-linear functions, and we have, © ain, 2), the vectors are i ; 2) 1g elements. Substituting Eq. (19.2) in f, ned 4. (19.5) ey In Eqs. (19.1) and (19, | controller does not contain any energy stor” we (Cur) Substituting Eq. (19.4) in Eq. (19.1), we Bet ee flr glCn) = (19.5) can be written as (xr) Ifthe input vector ris assumed to be constant, Fs refine a) ca single output system as shown in Fig. 19.2 t0 obtain the state space equa Let us tak given system. Fig. 19.2 Single output system | For plant we can write 1 Cine vv) me This gives the differential equation we ce 2c Puming cms ant t= 4 i, eun2ewu2e In state space form we can write the plant differential equations as follows = fiw and to nape nd ee The maximum effort of the controller ts +4, to respectively. Mathematically we can write where the function sgnir~t, ) has a value + 1 when (r ~ x,) 18 pos (r= x,) is negative In state space form we can write the controller differentual equations a {1° ne Fires and 4, e R(urde ode, eU sgn = Scanned with cateannerNon-ti 7 phase Plane Method inear Control Systems | 799 . ved that noneli beer observed i ‘near system m; a pS?) 1g approximations. We c; ‘ay be simplified _onable PI “an apply ver to a non-linear second-order system with @ \d-order Ply very easily a graph ; syste ove secon’ system. I; * graphical ‘ it oe ther existing Sete inva phase plane are a ea a the phase trajectory shes not appear explicitly in the Traian disadvantage in the has a ee nt tory for a second-ord jectory. Let us discuss a praphieal me Semin peste er system represented by graphical method for determining the send Hes b+ f(e.d)e+ £(de=r (19.7) aye and f,(c,d)e are funeti . pare KUO ) ‘ yi e functions of ¢ and ¢ respectively. puting %=¢ and % = 6, we get form Eq (19.7) 4 (19.8) ad (19.9) “flee )—fGemEt (19.10) Equation (19.10) gives the slope of the trajectory at any point (x, x,). Let us take (19.11) here m is constant. We can write F(x.X,)=™ (19.12) ibed by fhis curve is known as an isocline. se described by Eq. (19-12) and t rons eer line system as shown in Fig 19.3 with input r(t) = 2u(f) and nitial conditions: Any trajectory can cl Let us consider a se Cuput c(t) having the following i (0) 2-1 and (0)=0 Fom Fig, 19.2, we ean write _@—[aal 5 R (52 +25 +5)C(8) = SRO) - Taking ; aking inverse Laplace transform, W® get gis) Fi 198 second-order linear system 13) . €+2645e=5r=10 Scanned with CamScanner800 | contro! Systems Putting x, =eand x, = & comparing, Bq, (19.1.3) with Eq. (19.7) we po AA.) 2 Llyn, oS As per Fg, (19.13), we get The family of isoctines will have the following equation Sy 10 (nn (95 These lines intersect the x axis ata point having the following value at the x coordinate: The common point of all isoclines is (2, 0). If m= 0, the equation for the isoclines will be Hn eed site, Figure 19.4 shows the isoclines for different values of m, Togetthe phase plane: trajectory foranon-linear system, let us consider Fig. 19.5 where inductance Land capacitance C are linear but resistance R is non-linear. Let the current characteristics of the non-linear resistance be given by AV + BY? From Fig. 19.5, we get Fig. 19.4 Isoclines of Fig. 19.3 L v : — 0060" ~~ | 1 av 3 lve —_avapye zie Bydt-+C-— AV + BV) =0 (19.16) gn " ati E ome i Differentiating Eq, (19.16) wrt. t, we get av av tC ae Gy BIB? —aLjay op Fig. 19.5 A system Scanned with CamScannerNon-Linear Control Systems | 804 qeens take t ~ te av ave av (19.17) ude d ae Tie al av (i) (19.18) ae ic yar Using Eqs. (19.17) and (19.18), we get ftom Eq, (19.16) as SE e-abryer (19.19) d where and fe iE c 3B A Form Eq. (19.19), we have ay a Seting x,(t) = v(t) and x,(1) vite the followi Madore =e =1(t) and assuming the standard values @ = Zand b= 1, we can (19.20) Scanned with CamScanner802 | Control Systems Fig. 19.6 Isoclines and trajectories for a non-linear system Figure 19.6 shows the isoclines and trajectories for a non-linear system for different values of m. Fora non-linear system, we get a family of trajectories. Therefore, this diagram is also called phase portrait. 19.3 Describing Function Analysis The steady-state output of a non-linear device will have the same period of output, but there is 69 ‘guarantee that the output will be sinusoidal in spite of sinusoidal input. We know that the Fourier analysis of the fundamental wave form will lead to a fundamental component as well as several harmonics It is assumed that subharmonics do not exist. If the feedback system shown in Fig. 19.7 is of low pss nature, the harmonics will be attenuated and may be neglected to get a reasonable approximation to Performan: of the system. There are possibilities of oscillations if the loop gain is one with 2 piss shift of °. Non-lir A) —()- Gis en inear Saw Hy as Lo Hs) Fig. 19.7 Non-linear system Scanned with Camsénner“ratio of the findamental harmonic component of the out es Mpa ante cfement and i 1 en by al harmonic co of the input sinusoid, ¥, is the amplitude of the fundat s the phase shift of the fundamental harmonic componen' of th stpuit © input, Tain ef be the input to a non-Tinear clement, by Fourier series the output is Ive by 4+ DUA, sine +B, cos) = a + DY sin(not +0.) sie yan sin n ord (OD when non-linearity is symmetric. The fandamental component ofthe ourDut SS n(or+,)> y= Acoseat + B,sin oF where : LP ypsinor dion alo LE* yeneosordto” and ado The amplitude and phase angle of the fundamental component oF he output are given by (19.21) and (19.22) respectively. 19.3.1 Dead-Zone Non-Linearity Figue 19.8 shows idealized characteristis 07 sinusoidal input. No output can be ‘obtained within ‘element. -tineaity having dead-zone and its fesponse due to the dead-zone amplitude for inputs to 8 dead-zone Scanned with CamScannerN@xn NON SNE 804 | contro! systems Fig. 19.8 Sinusoidal response of non-linearity with dead-zone Ineart) From Fig. 19.8 the output (1) for 0S ev Sm ts given by 0 for O«t ® vite RL mnor 3) for cre ot 0 for : © ‘The output function is an odd function and hence 4, terms will be abs contains only sine terms. The fundamental harmonic component of the output 8 § y,(2) = B sin ew where Ups = =f vensin adie) a Due to half-wave symmetry, we can write 4 aerl rysin aden) © 8 f K(X sin or = Aysin or rs Scanned with CamScannerNon-Linear Control Systems | 805 in" (AIX) * jo (19.24), we get 4XK ff wae (sinor -sinor,)sinord(t) 4xK\1fn__ (A)|_A Ay =—|=}=-sin! en .25) x (Hs 2 (Fh xy! (3) oe sedescribing function ofan element with the dead zones given by B, 4K 1 A) A ay ot 20 =| X-— 1-0 | —|-— | S oo (19.26) Ney 5 {bs ‘) z 1 (2) 0 ) Figure 19.9 shows the plot of N/K versus $ : 10 08 06 ale 04 02 ws o 2 04 08 08 10 xI> Fig. 19.9 Describing function for the dead-zone non-linearity 19.3.2 Saturation Non-Linearity Figure 19.10 shows atypical input-output characteristic for he sarurtion non-linearity. In this case, the output of the saturation element is proportional to the input for small input signals, the output will not increase proportionally for larger input signals and at last the output is constant for very large signals. Treat 1510 shows that the ouput i odd and te fundamental component contains only sine terms. ‘The fundamental harmonic component of ouput is given By y(t) = B,sinor a, Scanned with CamScanner806 | Control Systems Output wy | KS. footy ks yall) = ¥1 sin ot x(t) =X sin ot ot Fig. 19.10 Sinusoidal response of non-linearity with saturation non-linearity where Jf xOsinord(on = 4p y(t)sinord (at) (97 © From Fig. 19.10, the output y(t) for 0< @r Using Eq, (19.28), we get form Eq, (19.27) 4 2 B. Ji v()sin wt d(wt) ™ = {faces ord(or) +f KSsinot avo] OS Scanned with CamScannerNon-Linear Control Systems | 807 +B+ seo] (19.29) i 08 where B=sin" (SIX). ‘The describing function of such an elementis given by. 08 N B, * y=—Z0° N XY 04 ak Jeni S}+$ 1-(5) | 0» 7 aa; Vt¥ Z} [20 09:30) 0 02 04 06 O08 1.0 The plot of N/K as a function of S/X is shown in 4 Fig. 19.11. * The relation between the describing function for Fig- 19.11 Describing function for the the dead zone non-linearity and that of saturation non- saturation non-linearity linearity is given by Neate = K-Neccine fOr = S (19.31) Output 19.3.3 On-Off Non-Linearity with Dead Zone Figure 19.12 shows the describing function for the on-off non-linearity with dead zone. “4 Figure 19.13 shows the input-output characteristics for the wt input on-off non-linearity with dead zone, It shows that the output is odd and the fundamental contains only sine terms. ; Form Fig. 19.13, the output (1) for $ wr <7 is given by Fig. 19.12 On-off non-linearity 0 ford (n-a) 0 for (n-0)2 012% Since the output is an odd term, WO= DB, cos cr. vt The output has also half-wave as well as quarter-wave symmetry also, The fundamental harmonic component is given by y(O= B, sino Fig. 19.16 Describing function for the on-off non-linearity x=Xsin ot Fig. 19.17 Sinusoidal response of non-linearitv witg, dead zone and sature' tion canned with CamScanner811 Non-Linear Control Systems: te an . 4 f wiesinen d(o)== [2 wsinotd(ot) . ; : K(Xsinor—aysinord(ony+ 3 [2 KS Aaysinord(o) x KX = P2B~ a) + (sin 2B —sin20)] she describing function is given by 0 for|X]S forsaturation non-linearity A=0 and a=0 Therefore, the describing function in given by | ; 1 for|X| ©, B= 7/2. Therefore, the describing function is given by 0 for|X|a Gane) hi "tre 1 '9.19 shows the describing function with dead-zone non-linearity N Ro 2 1-2 (.+sin cosa) = — — Scanned with CamScannero124 6 8 10 Fig. 19.18 Describing function of i Fig. 19.19 Describing func saturation non-linearity ng function ct dead-zone non ines For relay with dead zone In this cass ¢ S=A and slope K =< and the saturation value is |M|,Tasy a=B. the expression for N/M is given by : for|X|. 49.3.6 On-Off Non-Linearity with Hysteresis Figure 19.20 shows the input-output characteristic curve ofan on-off element. Let the inputx(0)=! X> h where h is the dead zone of the on-off non-linearity with hy ‘steresis. The output is given by M fort, Sr (99) x= _ - —M te(E+)ses(Z +) ® © “ sine tes 1) Due to odd symmetry of the output, its fundamental consis only sin x i: - ‘The fundamental harmonic component is given by where or, = sin” where Scanned with CamScannerNon-Linear Control Systems | 84 3 els a att eter x) = Xsin ot Y--f-----t- Fig. 19.20 Sinusoidal response of relay with hysteresis-type non-linearity The output is a square wave and it lags behind the input by an an; ‘ation for the non-linear element is given by 4M (4) = £-sin aX : sin"! (/X) and the describing Geseraly it is convenient to plot (A/M)N -( ad of N versus h/X. \é-sin ‘th t)versue h x 619.21 shows. the plot of AN/M versus AX’ iy : , 37 Relay with Dead Zone and Hysteresis "2219.29 sho ‘ben by Ws the input-output characteristics of a relay with dead-zone hysteresis. The output (1) 0 forosarsn +M for 0Serst way={0 for (x-B)Surs(x+ay =M_ for (n+a)S ors (2x-B) 0 for(2x-B) Sars 2e 8S sins Sm (AX) and B = sin’! [A-2h/1'] Scanned with CamScannerNon-Linear Control Systems | 817 2 al ==f3(xsina 2 2 ah 2 eos d(ary+ = fe "(xbewor avon : +2)" (xsinars? x zeal 5 Jeosex diary = -* cos: (19.41) ® a Qe 8, == J) sosinor dear) 22 pfxs 6 Qype b 2 J; [Xsiner 5} deo +2 () 3 mor duo +2 (sina +8 sino de =2[(E sp). sinay 942 al. m= 4 op )+ sins (19.42) Therefore, 0 for xj Vr N 4 2 Figure 19.25 shows the magnitude and phase characteristics of the describing function of a backlash see-linearity 19.4 Stability The stability charactenstic of a non-linear system abe determined using Lyapunov method The fol- lowing definitions are important before studying the os Lapunoy method. Lets consider a scalar function F of state variable Aig, Fla) = F(x, Xp ty o> Fe) If the functor has always a positive or negative sign 1 2 FEBS or the ongin except at the ongin, the funcbon ' ‘st be postive or negative definite The fsnctor Sessa is postive definite in OHSS Foca space except at the org! The functor “x? = 5x2 ay negative definite three aa space except at the origin. The funcen i+ (,

0_ for all x40, and dF(x) dt stem is asymptotically stable in the neighbourhood of the origin. On the other hang y < forall x#0 Then, the sy F(x) < g,, the system is stable but there is no assurance for asymptotic stability. dt known as Lyap Function Fi, unov’s function, Let us consider and autonomous system expressed as =m,x,—m,%, Let us take Lyapunov’s function as F et Here F = x? +x2 is positive definite. Therefore, (0) = 0. Now, dF _ OF 9x, OF ax; dt x, af Ox, a = Dai, + 2x, (-m.x, —m,x,) = 2x, —2xym,x, — 2M = (x, —m,x, —m,x,)x, = [201m )x, -2m,x,}y, For m,=1 and m,>0, we get 2 a-2m% | Therefore, for any value of x,,dF/dt = 19.4.2 Lyapunov’s Second Method IfA is a positive matrix, we can write if x, =0. F(x)= x" Ax oF () 8 et aet ata & Again = Br. Therefore, ¢? = x7 Br F(x) By TE Ant aT AR = xB Ax txt ABE = x1 (BTA+ ABE | / Scanned with CamScannerNon-Linear Control Systems \ 819 C=~(B'A+ 4B) we = (x)=, ref selecting ese definite or semi-definite matrix for C, we can solve for C. The neces- yn sicint condition for asympote stability is that he mati C must be postive definite ‘Any matric 1S Positive definite if its symmetric component is positive definite. Now 1 1 =(BTA+ +-(A" [s ‘A+ AB) 34 aan] - f(St}(44 )s| 2 2 2 Therefore, we can write = C = -(B'D + DA] where D is symmetric Let us consider an autonomous system expressed as Let C=(B'D+ DB)=1 ~ pale ave sls se From the above equation, we can write (19.43) (9.44) (19.45) Scanned with CamScanner820 | Control Systems From Eqs. (19-43) 19.45) we get To make D positive definite Dy, >0 Pu? (9% | and =D >0 DD, - Di, > ¢ ve From Eqs. (19.46) and (19.47) we get ae (9a and i [mi +(m, 1 J>0 To make Eg, (19.49) true, m, >0 and [{7m} + (m, +1)}/#t;] is greater than zero, Here m, >0 and to make D positive definite, m, must be greater than 2=10 For stabi method, m, >0 and m, > 0. In Lyapunov’s fi ust be greater than zero. Therefore, Lyapunov's first mett Lyapunov’s secon m, > 0 and m, mi SIGNIFICANT POINTS Non-Linear Control Systems: There are control systems which do pot have any difficult to linearize these control systems. In some cases tt has been fou non ‘and hysteresis and combinations of these are undesirable and unavoxtable linearized Dead-Zone Non-Linearity Saturation-Non-Linearity 8, Net so’= x On-Off Non-Linearity with Dead Zone r New S (A Jo Cs Scanned with CamScanner4 Non-Linear Control Systems | 82 on-Linearity om P y Narco’ coe x me pone aa sacaration 1 x e for|X| 8 usatration Non-Linearity D=0 and a=0 ‘erfre, the describing funetion is given by 1 for |X |< for |X|25 Dead-Zone Non-linearit lathis ease § — oo, Therefore, the describing function is given by for|xl A n value is |M|. Taking a&=B, the expression for N/M is 2 = (atsinacosa m 'erRelay with Dead Zone this case $= A and slope K =e> and the saturatios Brady for|X|< A. for|x|>4 r Scanned with CamScanner822 | contro! systems Relay with Dead-Zone Hysteresis 4y (2) for|x|>4 wet\x) “Ur 4 0 for|X}< a Backlash o foe xj = Lyapunov’s First Theorem: If there exists 2 salar function F(x) for an autonomous system t= f(x) such tat F(x) is real, continuous with continuous first partial derivatives and FO=0 Fix)>0 forall e#0 dFay a <0 forall seo Then, the system is asymptotically stable in the merghhonsrt of the origin. On the the system is stable but there is no assurance for function. Lyapunoy's Second Method: If 4 1s 2 positive matrix, we ean write ryimptotic stabelity The function Fx) is known as Lyapunov’s Fiestas ary & =U Areas Again x= Bx Therefore x’ = xB", ain Ae dee eB dae ABN BAe AB) & Let Cm (BA AB) ay Therefore, 8 Fue Cx Selecting arbitrarily a positive definite or semidetinite matetx for C, we can solve for © TH sufficient condition for asymptotic stability x that the matrix C rmust he positive detinite uA Scanned with CamScanner