Classical Optimization Techniques: Hapter
Classical Optimization Techniques: Hapter
f(x)
B1 a b x a b x
53
A single-variable optimization problem is one in which the value of x = x* is to be found in the interval [a, b] such that x* minimizes f (x). The following two theorems provide the necessary and sufficient conditions for the relative minimum of a function of a single variable. Theorem 2.1: Necessary Condition If a function f (x) is defined in the interval a x b and has a relative minimum at x = x*, where a < x* < b, and if the derivative df (x)/dx = f (x) exists as a finite number at x = x*, then f(x*) = 0. Proof: It is given that
f ( x * ) = lim f ( x* + h ) f ( x* ) h
h0
...(2.1)
exists as a definite number, which we want to prove to be zero. Since x* is a relative minimum, we have
f ( x* ) f ( x* + h ) for all values of h sufficiently close to zero. Hence
f ( x* + h ) f ( x* ) 0 h
if h > 0
f ( x* + h ) f ( x* ) 0 if h < 0 h Thus Eq. (2.1) gives the limit as h tends to zero through positive values as
f ( x* ) 0
...(2.2)
...(2.3)
The only way to satisfy both Eqs. (2.2) and (2.3) is to have
f ( x* ) = 0
...(2.4)
This proves the theorem. Notes: 1. This theorem can be proved even if x* is a relative maximum. 2. The theorem does not say what happens if a minimum or maximum occurs at a point x* where the derivative fails to exist. For example, in Fig. 2.2,
h 0
lim
f ( x * + h ) f ( x* ) = m + ( positive ) or m ( negative ) h
depending on whether h approaches zero through positive or negative values, respectively. Unless the numbers m+ and m are equal, the derivative f (x*) does not exist. If f (x*) does not exist, the theorem is not applicable. 3. The theorem does not say what happens if a minimum or maximum occurs at an endpoint of the interval of definition of the function. In this case
AND
PRACTICE
f ( x* + h ) f ( x* ) h 0 h exists for positive value of h only or for negative value of h only, and hence the derivative is not defined at the endpoints. lim
f(x) Negative slope m
-
m f(x*)
Positive slope m
x x*
4. The theorem does not say that the function necessarily will have a minimum or maximum at every point where the derivative is zero. For example, the derivative f (x) = 0 at x = 0 for the function shown in Fig. 2.3. However, this point is neither a minimum nor a maximum. In general, a point x* at which f (x*) = 0 is called a stationary point. If the function f (x) possesses continuous derivatives of every order that come in question, in the neighbourhood of x = x*, the following theorem provides the sufficient condition for the minimum or maximum value of the function. Theorem 2.2: Sufficient Condition. Let f (x*) = f (x*) = = f (n1) (x*) = 0, but f (n) (x*) 0. Then f (x*) is (i) a minimum value of f (x) if f (n)(x*) > 0 and n is even; (ii) a maximum value of f (x) if f (n) (x*) < 0 and n is even; (iii) neither a maximum nor a minimum if n is odd.
55
f ( x * + h ) = f ( x * ) + hf ( x * ) +
h 2 ( * ) f x + 2!
h n 1 ( n1) ( * ) f x ( n 1)!
...(2.5)
+
Since
f ( x* + h ) f ( x* ) =
hn (n) ( * f x + h ) n!
As f (n)(x*) 0, there exists an interval around x* for every point x of which the nth derivative f (n)(x) has the same sign, namely, that of f (n)(x*). Thus for every point x* + h of this interval, f (n)(x* + h) has the sign of f (n)(x*). When n is even, hn/n! is positive irrespective of whether h is positive or negative, and hence f (x* + h) f (x*) will have the same sign as that of f (n)(x*). Thus x* will be a relative minimum if f (n)(x*) is positive and a relative maximum if f (n)(x*) is negative. When n is odd, hn/n! changes sign with the change in the sign of h and hence the point x* is neither a maximum nor a minimum. In this case the point x* is called a point of inflection. Example 2.1: Determine the maximum and minimum values of the function f (x) = 12x5 45x4 + 40x3 + 5 Solution: Since f (x) = 60(x4 3x3 + 2x2) = 60x2(x 1) (x 2), f (x) = 0 at x = 0, x = 1, and x = 2. The second derivative is f (x) = 60(4x3 9x2 + 4x) At x = 1, f (x) = 60 and hence x = 1 is a relative maximum. Therefore, fmax = f (x = 1) = 12 At x = 2, f (x) = 240 and hence x = 2 is a relative minimum. Therefore, fmin = f (x = 2) = 11 At x = 0, f (x) = 0 and hence we must investigate the next derivative. f (x) = 60(12x2 18x + 4) = 240 at x = 0 Since f (x) 0 at x = 0, x = 0 is neither a maximum nor a minimum, and it is an inflection point. Example 2.2: In a two-stage compressor, the working gas leaving the first stage of compression is cooled (by passing it through a heat exchanger) before it enters the second stage of compression to increase the efficiency [2.13]. The total work input to a compressor (W) for an ideal gas, for isentropic compression, is given by
p ( k 1) k p ( k 1) k W = c pT1 2 + 3 2 p1 p2
where cp is the specific heat of the gas at constant pressure, k is the ratio of specific heat at constant pressure to that at constant volume of the gas, and T1 is the temperature at which the gas enters the compressor. Find the pressure, p2, at which intercooling should be done to minimize the work input to the compressor. Also determine the minimum work done on the compressor.
AND
PRACTICE
Solution: The necessary condition for minimizing the work done on the compressor is:
( k 1) k dW k 1 k 1 1 k = c p T1 ( p2 ) dp2 k 1 p1 k
+ ( p3 ) which yields
( k 1) k
(1 2 k ) k k + 1 ( p2 ) =0 k
p2 = ( p1 p3 )
12
d 2W 2 dp 2
k 1 2c p T1 = ( 3 k 1) 2 k (k +1) 2 k 12 p1 p3k p2 =( p1 p2 )
and hence the solution corresponds to a relative minimum. The minimum work done is given by
Wmin = 2c pT1
( k 1) 2 k k p3 1 k 1 p1
i =1 j =1
...
k =1
hi h j ... hk
r f ( X* ) xi x j ... x k
...(2.6)
r summations
is called the rth differential of f at X*. Notice that there are r summations and one hi is associated with each summation in Eq. (2.6). For example, when r = 2 and n = 3, we have
57
h h
i i =1 j =1
2 f ( X* ) x i x j
2 = h1
2 2 2 f ( * ) 2 f 2 f X + h2 2 ( X * ) + h3 2 ( X * ) 2 x1 x 2 x3
+ 2 h1h2
The Taylors series expansion of a function f (X) about a point X* is given by f ( X ) = f ( X * ) + df ( X * ) + + + 1 2 ( *) 1 3 ( *) d f X + d f X 2! 3! ...(2.7)
1 N ( *) d f X + RN ( X* , h ) N!
( N + 1)!
d N +1 f ( X * + h )
...(2.8)
Example 2.3: Find the second-order Taylors series approximation of the function
2 f ( x1 , x2 , x3 ) = x 2 x3 + x1e x 3
Solution: The second-order Taylors series approximation of the function f about point X* is given by
1 f ( X ) = f 0 + df 0 + d 2 f 0 2! 2 2 2 1 f 0 = e 2 2 1 1 1 0 = h f 0 + h f 0 + h f df 1 2 3 x1 x 2 x 3 2 2 2 1 0 2 1 1 1
where
AND
PRACTICE
1 2 2 2 0 = h 2 f + h 2 f + h 2 f 2 3 2 2 1 x 2 x 2 x 3 1 2 1 0 2
1 h 2 ( 0 ) + h 2 ( 2 x ) + h 2 ( x e x 3 ) + 2h h ( 0 ) + 2 h h ( 2 x ) + 2h h ( e x 3 ) 0 = 1 2 3 3 1 1 2 2 3 2 1 3 2
2 2 = 4 h2 + e 2 h3 + 2h1h3 e 2
Thus the Taylors series approximation is given by f (X) e 2 + e 2 ( h1 + h3 ) + 1 ( 4h22 + e 2 h32 + 2h1h3e 2 ) 2!
where h1 = x1 1, h2 = x2, and h3 = x3 + 2. Theorem 2.3: Necessary Condition. If f (X) has an extreme point (maximum or minimum) at X = X* and if the first partial derivatives of f (X) exist at X*, then
f ( * ) f ( * ) f ( * ) X = X = ... = X =0 x1 x2 x n
...(2.9)
Proof: The proof given for Theorem 2.1 can easily be extended to prove the present theorem. However, we present a different approach to prove this theorem. Suppose that one of the first partial derivatives, say the kth one, does not vanish at X*. Then, by Taylors theorem,
f ( X* + h ) = f ( X * ) +
f ( X * + h ) f ( X * ) = hk
h x
i i =1
( X* ) + R1 ( X* , h )
i
that is,
f ( * ) 1 2 ( * X + d f X + h ) , x k 2!
0<<1
Since d2f (X* + h) is of order hi2 , the terms of order h will dominate the higher-order terms for small h. Thus the sign of f (X* + h) f (X*) is decided by the sign of hk f (X*)/xk. Suppose that f (X*)/xk > 0. Then the sign of f (X* + h) f (X*) will be positive for hk > 0 and negative for hk < 0.
59
This means that X* cannot be an extreme point. The same conclusion can be obtained even if we assume that f (X*)/xk < 0. Since this conclusion is in contradiction with the original statement that X* is an extreme point, we may say that f/xk = 0 at X = X*. Hence the theorem is proved. Theorem 2.4: Sufficient Condition. A sufficient condition for a stationary point X* to be an extreme point is that the matrix of second partial derivatives (Hessian matrix) of f (X) evaluated at X* is (i) positive definite when X* is a relative minimum point, and (ii) negative definite when X* is a relative maximum point. Proof: From Taylors theorem we can write
f ( X* + h ) = f ( X* ) +
i =1
hi
f ( * ) 1 X + xi 2!
i =1 j =1
hi h j
2 f xi x j
X = X * +h
0<<1 Since X* is a stationary point, the necessary conditions give (Theorem 2.3)
f = 0, i = 1, 2, ..., n x i
...(2.10)
i =1 j =1
hi h j
2 f xi x j
X = X* +h
, 0<<1
i =1 j =1
hi h j
2
2 f x i x j
X = X * +h
Since the second partial derivative of f (X)/xi xj is continuous in the neighborhood of X*,
2 f xi x j
2
X = X* +h
will have the same sign as ( f/xi xj)|X = X* for all sufficiently small h. Thus f (X* + h) f (X*) will be positive, and hence X* will be a relative minimum, if
Q=
i =1 j =1
hi h j
2 f xi x j
...(2.11)
X = X*
is positive. This quantity Q is a quadratic form and can be written in matrix form as Q = h T JH X =X*
2 f J X = X* = where xi x j X = X* is the matrix of second partial derivatives and is called the Hessian matrix of f (X).
...(2.12)
...(2.13)
AND
PRACTICE
It is known from matrix algebra that the quadratic form of Eq. (2.11) or (2.12) will be positive for all h if and only if [J] is positive definite at X = X*. This means that a sufficient condition for the stationary point X* to be a relative minimum is that the Hessian matrix evaluated at the same point be positive definite. This competes the proof for the minimization case. By proceeding in a similar manner, it can be proved that the Hessian matrix will be negative definite if X* is a relative maximum point. Note: A matrix A will be positive definite if all its eigenvalues are positive; that is, all the values of that satisfy the determinantal equation |A I| = 0 ...(2.14) should be positive. Similarly, the matrix [A] will be negative definite if its eigenvalues are negative. Another test that can be used to find the positive definiteness of a matrix A of order n involves evaluation of the determinants A = |a11|,
A2 =
a11 a21
a12 a22
a12 a22 a32
,
a13 a23 ..., a33
a1n a2 n a3 n , ann
...(2.15)
The matrix A will be positive definite if and only if all the values A1, A2, A3, ..., An are positive. The matrix A will be negative definite if and only if the sign of Aj is (1)j for j = 1, 2,..., n. If some of the Aj are positive and the remaining Aj are zero, the matrix A will be positive semidefinite. Example 2.4: Figure 2.4 shows two frictionless rigid bodies (carts) A and B connected by three linear elastic springs having spring constants k1, k2, and k3. The springs are at their natural positions when the
k1 B
k2
A k3
x1
x2
61
applied force P is zero. Find the displacements x1 and x2 under the force P by using the principle of minimum potential energy. Solution: According to the principle of minimum potential energy, the system will be in equilibrium under the load P, if the potential energy is a minimum. The potential energy of the system is given by potential energy (U) = strain energy of springs work done by external forces 1 1 2 1 2 2 = k2 x1 + k3 ( x 2 x1 ) + k1 x 2 Px 2 2 2 2 The necessary conditions for the minimum of U are
U = k2 x1 k3 ( x 2 x1 ) = 0 x1 U = k3 ( x 2 x1 ) + k1 x2 P = 0 x 2
...(E1)
...(E2)
The values of x1 and x2 corresponding to the equilibrium state, obtained by solving Eqs. (E1) and (E2) are given by
* x1 =
Pk 3 k1k 2 + k1k3 + k 2 k 3
P ( k 2 + k3 ) k1k 2 + k1k 3 + k 2 k3
* x2 =
2U 2 x1 * J ( x1* , x2 ) = 2 U x1x2
2U x1x2 k + k3 = 2 2 U k3 2 x2 ( x* , x* ) 1 2
k3 k1 + k3
since the spring constants are always positive. Thus the matrix J is positive definite and hence (x*, x*) 1 2 corresponds to the minimum of potential energy.
AND
PRACTICE
2.3.1 Semidefinite Case We now consider the problem of determining the sufficient conditions for the case when the Hessian matrix of the given function is semidefinite. In the case of a function of a single variable, the problem of determining the sufficient conditions for the case when the second derivative is zero was resolved quite easily. We simply investigated the higher-order derivatives in the Taylors series expansion. A similar procedure can be followed for functions of n variables. However, the algebra becomes quite involved, and hence we rarely investigate the stationary points for sufficiency in actual practice. The following theorem, analogous to Theorem 2.2, gives the sufficient conditions for the extreme points of a function of several variables. Theorem 2.5: Let the partial derivatives of f of all orders up to the order k 2 be continuous in the neighborhood of a stationary point X*, and
dr f
dk f
X = X
= 0,
0,
1rk1
X =X
so that d k f
X = X*
is a relative minimum if d k f
X = X*
is
X = X*
drawn. On the other hand, if k is odd, X* is not an extreme point of f (X). Proof: A proof similar to that of Theorem 2.2 can be found in Ref. [2.5]. 2.3.2 Saddle Point In the case of a function of two variables, f (x, y), the Hessian matrix may be neither positive nor negative definite at a point (x*, y*) at which f f = =0 x y In such a case, the point (x*, y*) is called a saddle point. The characteristic of a saddle point is that it corresponds to a relative minimum or maximum of f (x, y) with respect to one variable, say, x (the other variable being fixed at y = y*) and a relative maximum or minimum of f (x, y) with respect to the second variable y (the other variable being fixed at x*). As an example, consider the function f (x, y) = x2 y2. For this function, f f = 2 x and = 2 y x y These first derivatives are zero at x* = 0 and y* = 0. The Hessian matrix of f at (x*, y*) is given by
2 0 J= 0 2
Since this matrix is neither positive definite nor negative definite, the point (x* = 0, y* = 0) is a saddle point. The function is shown graphically in Fig. 2.5. It can be seen that f (x, y*) = f (x, 0) has a
63
relative minimum and f (x*, y) = f (0, y) has a relative maximum at the saddle point (x*, y*). Saddle points may exist for functions of more than two variables also. The characteristic of the saddle point stated above still holds provided that x and y are interpreted as vectors in multidimensional cases.
f(x, y)
(x*, y*) x
Solution: The necessary conditions for the existence of an extreme point are
f 2 = 3 x1 + 4 x1 = x1 ( 3 x1 + 4 ) = 0 x1
f 2 = 3 x2 + 8 x 2 = x 2 ( 3 x2 + 8 ) = 0 x1
( 0, 0 ) , 0,
8 , 3
8 4 4 , 0 and , 3 3 3
To find the nature of these extreme points, we have to use the sufficient conditions. The secondorder partial derivatives of f are given by
2 f = 6 x1 + 4 2 x1 2 f = 6 x2 + 8 2 x2
AND
PRACTICE
2 f =0 x1x 2
0 6 x + 4 J= 1 6 x2 + 8 0
If J1 = |6x1 + 4| and J 2 = point are as given below.
Value Point X (0, 0) of J1 +4 +4 4 4 Value of J2 +32 32 32 Nature of J Positive definite Indefinite Nature of X Relative minimum Saddle point f (X) 6 418/27
8 0, 3 4 , 0 3 4 8 , 3 3
Indefinite
Saddle point
194/27
+32
Negative definite
Relative maximum
50/3
where
x1 x X = 2 xn
Here m is less than or equal to n; otherwise (if m > n), the problem becomes overdefined and, in general, there will be no solution. There are several methods available for the solution of this problem. The methods of direct substitution, constrained variation, and Lagrange multipliers are discussed in the following sections. 2.4.1 Solution by Direct Substitution For a problem with n variables and m equality constraints, it is theoretically possible to solve
65
simultaneously the m equality constraints and express any set of m variables in terms of the remaining n m variables. When these expressions are substituted into the original objective function, there results a new objective function involving only n m variables. The new objective function is not subjected to any constraint, and hence its optimum can be found by using the unconstrained optimization techniques discussed in Section 2.3. This method of direct substitution, although it appears to be simple in theory, is not convenient from practical point of view. The reason for this is that the constraint equations will be nonlinear for most of practical problems, and often, it becomes impossible to solve them and express any m variables in terms of the remaining n m variables. However, the method of direct substitution might prove to be very simple and direct for solving simpler problems, as shown by the following example. Example 2.6: Find the dimensions of a box of largest volume that can be inscribed in a sphere of unit radius. Solution: Let the origin of the Cartesian coordinate system x1, x2, x3 be at the center of the sphere and the sides of the box be 2x1, 2x2, and 2x3. The volume of the box is given by f (x1, x2, x3) = 8x1x2x3 ...(E1) Since the corners of the box lie on the surface of the sphere of unit radius, x1, x2, and x3 have to satisfy the constraint
2 2 2 x1 + x2 + x3 = 1
...(E2)
This problem has three design variables and one equality constraint. Hence the equality constraint can be used to eliminate any one of the design variables from the objective function. If we choose to eliminate x3, Eq. (E2) gives
2 2 x3 = (1 x1 x2 ) 12
...(E3)
12
...(E4)
which can be maximized as an unconstrained function in two variables. The necessary conditions for the maximum of f give
2 12 x1 f 2 2 = 8 x2 (1 x1 x2 ) =0 12 x1 (1 x12 x22 ) 2 12 x2 f 2 2 = 8 x1 (1 x1 x2 ) =0 12 2 2 x 2 (1 x1 x2 ) Equations (E5) and (E6) can be simplified to obtain
2 2 1 2 x1 x 2 = 0 2 2 1 x1 2 x 2 = 0 * from which it follows that x1 = x * = 1 2 volume of the box as * 3 and hence x3 = 1
...(E5)
...(E6)
AND
PRACTICE
fmax =
8 3 3
To find whether the solution found corresponds to a maximum or a minimum, we apply the sufficiency conditions to f (x1, x2) of Eq. (E4). The second-order partial derivatives of f at given by
3 12 8 x1 x2 8 x2 x1 2 f 2 2 = + 2 x1 (1 x1 x2 ) 2 12 2 2 12 x1 (1 x12 x22 ) 1 x1 x2 (1 x12 x22 )
( x1* , x2* )
are
32 3
* * at ( x1 , x 2 )
32 3
* * at ( x1 , x 2 )
2 12 x2 2 2 (1 x1 x 2 ) + 12 (1 x12 x22 )
= Since
16 3
* * at ( x1 , x 2 )
2 f 2 f 2 f 2 f < 0 and 2 2 2 x1 x1 x 2 x1 x 2
>0
* * * * the Hessian matrix of f is negative definite at ( x1 , x2 ) . Hence the point ( x1 , x2 ) corresponds to the
maximum of f. 2.4.2 Solution by the Method of Constrained Variation The basic idea used in the method of constrained variation is to find a closed-form expression for the first-order differential of f (df) at all points at which the constraints gj (X) = 0, j = 1, 2,..., m, are satisfied. The desired optimum points are then obtained by setting the differential df equal to zero. Before presenting the general method, we indicate its salient features through the following simple problem with n = 2 and m = 1. Minimize f (x1, x2) ...(2.17)
67
* of f (x1, x2) with respect to x1 must be zero at x1 , x* . By setting the total differential of f (x1, x2) equal 2
to zero, we obtain
df = f f dx1 + dx 2 = 0 x1 x2
...(2.19)
* * Since g ( x1 , x * ) = 0 at the minimum point, any variations dx1 and dx2 taken about the point ( x1 , x * ) 2 2 are called admissible variations provided that the new point lies on the constraint: * * g ( x1 + dx1 , x 2 + dx 2 ) = 0
...(2.20)
* The Taylors series expansion of the function in Eq. (2.20) about the point ( x1 , x * ) gives 2 * g ( x1 + dx1 , x * + dx2 ) 2
* * g ( x1 , x 2 ) +
...(2.21)
* where dx1 and dx2 are assumed to be small. Since g ( x1 , x * ) = 0, Eq. (2.21) reduces to 2
dg =
g g dx1 + dx2 = 0 at x1 x2
( x1* , x2* )
...(2.22)
Thus Eq. (2.22) has to be satisfied by all admissible variations. This is illustrated in Fig. 2.6, where PQ indicates the curve at each point of which Eq. (2.18) is satisfied. If A is taken as the base point
* ( x1* , x2 ) , the variations in x1 and x2 leading to points B and C are called admissible variations. On the
other hand, the variations in x1 and x2 representing point D are not admissible since point D does not
x2 Q C A B g(x1, x2) = 0 x1
lie on the constraint curve, g(x1, x2) = 0. Thus any set of variations (dx1, dx2) that does not satisfy Eq. (2.22) lead to points such as D which do not satisfy constraint Eq. (2.18).
AND
PRACTICE
dx 2 =
g x1 * * ( x1 , x2 ) dx1 g x 2
...(2.23)
This relation indicates that once the variation in x1 (dx1) is chosen arbitrarily, the variation in x2 (dx2) is decided automatically in order to have dx1 and dx2 as a set of admissible variations. By substituting Eq. (2.23) in Eq. (2.19), we obtain
g x1 f f df = dx1 = 0 * x1 g x2 x2 ( x1* , x2 )
...(2.24)
The expression on the left-hand side is called the constrained variation of f. Note that Eq. (2.24) has to be satisfied for all values of dx1. Since dx1 can be chosen arbitrarily, Eq. (2.24) leads to
f g f g =0 x x x x 2 1 ( x * , x* ) 1 2 1 2
* * Equation (2.25) represents a necessary condition in order to have x1 , x 2
...(2.25)
as an extreme point
(minimum or maximum). Example 2.7: A beam of uniform rectangular cross-section is to be cut from a log having a circular cross-section of diameter 2a. The beam has to be used as a cantilever beam (the length is fixed) to carry a concentrated load at the free end. Find the dimensions of the beam that correspond to the maximum tensile (bending) stress carrying capacity. Solution: From elementary strength of materials, we know that the tensile stress induced in a rectangular beam () at any fiber located a distance y from the neutral axis is given by M = y I where M is the bending moment acting and I is the moment of inertia of the cross-section about the x axis. If the width and depth of the rectangular beam shown in Fig. 2.7 are 2x and 2y, respectively, the maximum tensile stress induced is given by
max =
M My 3 M = y= 3 1 ( ) I 4 xy 2 2 x (2 y ) 12
Thus for any specified bending moment, the beam is said to have maximum tensile stress carrying capacity if the maximum induced stress (max) is a minimum. Hence we need to minimize k/xy2 or maximize Kxy2, where k = 3M/4 and K = 1/k, subject to the constraint x2 + y2 = a2 This problem has two variables and one constraint; hence Eq. (2.25) can be applied for finding the optimum solution. Since f = kx1y2 ...(E1)
69
a 2y x (Neutral axis)
x +y =a
2x
g = x2 + y2 a2 we have
f = kx 2 y 2 x
...(E2)
f = 2 kx 1 y 3 y
g = 2x x
g = 2y y Equation (2.25) gives that is, kx2y2(2y) + 2kx1 y3 (2x) = 0 at (x*, y*) y* = 2 x * ...(E3)
Thus the beam of maximum tensile stress carrying capacity has a depth of The optimum values of x and y can be obtained from Eqs. (E3) and (E2) as x* = a and y * = 2 a
3 3 Necessary Conditions for a General Problem. The procedure indicated above can be generalized to the case of a problem in n variables with m constraints. In this case, each constraint equation gj (X) = 0, j = 1, 2,..., m, gives rise to a linear equation in the variations dxi, i = 1, 2,..., n. Thus there will be in all m linear equations in n variations. Hence any m variations can be expressed in terms of the remaining n m variations. These expressions can be used to express the differential of the objective function, df, in terms of the n m independent variations. By letting the coefficients of the independent