2–7 LINEARIZATION OF NONLINEAR MATHEMATICAL MODELS where the derivatives df dx, d2f dx 2, p are evaluated at x = x– .

If the variation x - x–
is small, we may neglect the higher-order terms in x - x– . Then Equation (2–43) may be
Nonlinear Systems. A system is nonlinear if the principle of superposition does written as
not apply. Thus, for a nonlinear system the response to two inputs cannot be calculated
by treating one input at a time and adding the results. y = y– + K(x - x– ) (2–44)
Although many physical relationships are often represented by linear equations,
in most cases actual relationships are not quite linear. In fact, a careful study of phys- where
ical systems reveals that even so-called “linear systems” are really linear only in lim-
y– = f(x– )
ited operating ranges. In practice, many electromechanical systems, hydraulic systems,
pneumatic systems, and so on, involve nonlinear relationships among the variables.
df
For example, the output of a component may saturate for large input signals. There may K = 2
be a dead space that affects small signals. (The dead space of a component is a small dx x = x–
range of input variations to which the component is insensitive.) Square-law nonlin-
earity may occur in some components. For instance, dampers used in physical systems Equation (2–44) may be rewritten as
may be linear for low-velocity operations but may become nonlinear at high veloci-
ties, and the damping force may become proportional to the square of the operating
y - y– = K(x - x– ) (2–45)
velocity.

Linearization of Nonlinear Systems. In control engineering a normal operation which indicates that y - y– is proportional to x - x– . Equation (2–45) gives a linear math-
of the system may be around an equilibrium point, and the signals may be considered ematical model for the nonlinear system given by Equation (2–42) near the operating
small signals around the equilibrium. (It should be pointed out that there are many ex- point x = x– , y = y– .
ceptions to such a case.) However, if the system operates around an equilibrium point Next, consider a nonlinear system whose output y is a function of two inputs x1 and
and if the signals involved are small signals, then it is possible to approximate the non- x2 , so that
linear system by a linear system. Such a linear system is equivalent to the nonlinear sys-
tem considered within a limited operating range. Such a linearized model (linear, y = fAx1 , x2 B (2–46)
time-invariant model) is very important in control engineering.
The linearization procedure to be presented in the following is based on the ex- To obtain a linear approximation to this nonlinear system, we may expand Equation (2–46)
pansion of nonlinear function into a Taylor series about the operating point and the into a Taylor series about the normal operating point x– 1 , x– 2 . Then Equation (2–46)
retention of only the linear term. Because we neglect higher-order terms of the Taylor becomes
series expansion, these neglected terms must be small enough; that is, the variables
deviate only slightly from the operating condition. (Otherwise, the result will be 0f 0f
inaccurate.) y = fAx– 1 , x– 2 B + c Ax1 - x– 1 B + Ax2 - x– 2 B d
0x1 0x2

Linear Approximation of Nonlinear Mathematical Models. To obtain a linear

mathematical model for a nonlinear system, we assume that the variables deviate only 1 0 2f 0 2f
c 2 Ax1 - x– 1 B + 2 Ax - x– 1 BAx2 - x– 2 B
2
+
slightly from some operating condition. Consider a system whose input is x(t) and out- 2! 0x 1 0x1 0x2 1
put is y(t). The relationship between y(t) and x(t) is given by
0 2f
Ax2 - x– 2 B d + p
2
y = f(x) (2–42) +
0x 22
If the normal operating condition corresponds to x– , y– , then Equation (2–42) may be
expanded into a Taylor series about this point as follows: where the partial derivatives are evaluated at x1 = x– 1 , x2 = x– 2 . Near the normal oper-
ating point, the higher-order terms may be neglected. The linear mathematical model of
y = f(x) this nonlinear system in the neighborhood of the normal operating condition is then
given by
df 1 d 2f
= f(x– ) + (x - x– ) + (x - x– )2 + p (2–43) y - y– = K1 Ax1 - x– 1 B + K2 Ax2 - x– 2 B
dx 2! dx2

Section 2–7 / Linearization of Nonlinear Mathematical Models 43 44 Chapter 2 / Mathematical Modeling of Control Systems
 where EXAMPLE PROBLEMS AND SOLUTIONS
y– = fAx– 1 , x– 2 B
A–2–1. Simplify the block diagram shown in Figure 2–17.
0f
K1 = 2
0x1 x1 = x– 1 , x2 = x– 2 Solution. First, move the branch point of the path involving H1 outside the loop involving H2 , as
shown in Figure 2–18(a). Then eliminating two loops results in Figure 2–18(b). Combining two
0f blocks into one gives Figure 2–18(c).
K2 = 2
0x2 x1 = x– 1 , x2 = x– 2 A–2–2. Simplify the block diagram shown in Figure 2–19. Obtain the transfer function relating C(s) and
R(s).
The linearization technique presented here is valid in the vicinity of the operating
condition. If the operating conditions vary widely, however, such linearized equations are
H1
not adequate, and nonlinear equations must be dealt with. It is important to remember
that a particular mathematical model used in analysis and design may accurately rep-
resent the dynamics of an actual system for certain operating conditions, but may not be
R(s) + C(s)
accurate for other operating conditions. + G +
–

EXAMPLE 2–5 Linearize the nonlinear equation

Figure 2–17
z=xy Block diagram of a H2
in the region 5 x 7, 10 y 12. Find the error if the linearized equation is used to calcu- system.
late the value of z when x=5, y=10.
Since the region considered is given by 5 x 7, 10 y 12, choose x– = 6, y– = 11. Then
z– = x– y– = 66. Let us obtain a linearized equation for the nonlinear equation near a point x– = 6, H1
y– = 11. G
Expanding the nonlinear equation into a Taylor series about point x = x– , y = y– and neglecting
the higher-order terms, we have
R(s) + C(s)
z - z– = aAx - x– B + bAy - y– B (a) +
–
G +

where
0(xy)
a = 2 = y– = 11 H2
0x x = x– , y = y–

0(xy)
b = 2 = x– = 6
0y x = x– , y = y–
R(s) G H1 C(s)
(b) 1+
1 + GH2 G
Figure 2–18
Hence the linearized equation is
Simplified block
z-66=11(x-6)+6(y-11) diagrams for the
R(s) G + H1 C(s)
or system shown in (c)
Figure 2–17. 1 + GH2
z=11x+6y-66

When x=5, y=10, the value of z given by the linearized equation is

R(s) X(s) C(s)
z=11x+6y-66=55+60-66=49 G1 + G2 +
+ +
The exact value of z is z=xy=50. The error is thus 50-49=1. In terms of percentage, the Figure 2–19
error is 2%. Block diagram of a
system.

Section 2–7 / Linearization of Nonlinear Mathematical Models 45 46 Chapter 2 / Mathematical Modeling of Control Systems