Вы находитесь на странице: 1из 7

OPTIMIZATION OF CHEMICAL

PROCESSES (CHE1011)

Dr. Dharmendra Kumar Bal


Assistant Professor (Sr.)
School of Chemical Engineering

11
Polynomial Approximation Methods
These methods are developed assuming convex objective function.
They have mainly two steps:

(a) Approximate the nonlinear objective function by a suitable


polynomial, and

(b) Estimate the minimum using the polynomial,

The two steps have to be repeated until convergence.

Similar strategy is employed in some methods (e.g., successive


quadratic programming, SQP) for solving multivariable constrained
problems.
2
Polynomial Approximation Methods
A method based on quadratic approximation is discussed here.

Quadratic Approximation
Consider fitting a quadratic function:
qa(x) = a + b (x – x1) + c (x – x1) (x – x2)
to three points: (x1, f1); (x2, f2) and (x3, f3).

Coefficients in the quadratic are: a = f1; b = (f2 - f1) / (x2 - x1)


and c = [(f3 - f1) / (x3 - x1) – b] / (x3 – x2)

The minimum of qa(x) is at x = {(x1 + x2) / 2} – {b / (2 c)}

3
Polynomial Approximation Methods

Successive Quadratic Approximation Method

1. Select the initial point, x1 and step size, x. Evaluate f1.

2. Find x2 = x1 + x and f2.

3. If f1 > f2, then let x3 = x2 + 2 x. Else, x3 = x1 – x. Compute f3.

4. Determine fm = minimum of (f1, f2, f3), and the corresponding xm.

4
Polynomial Approximation Methods

5. Find x using the 3 points: (x1, f1), (x2, f2) and (x3, f3), and the quadratic
approximation formula.

Coefficients in the quadratic are: a = f1; b = (f2 - f1) / (x2 - x1)


and c = [(f3 - f1) / (x3 - x1) – b] / (x3 – x2)
Minimum of qa(x) is at x = {(x1 + x2) / 2} – {b / (2 c)}

6. Check for convergence. If abs[fm – f(x)] and/or abs(xm – x) are


acceptably small, then terminate. Else, continue to next step.

7. Save the currently best point (xm or x) and the two points bracketing
it. Label them as x1, x2 and x3, and repeat from step 4.
5
Polynomial Approximation Methods

Example: Minimize f(x) = x2 + 50 / x


Starting from x0 = 1.0 and using x = 1.

Quadratic Approximation: qa(x) = a + b (x – x1) + c (x – x1) (x – x2)


a = f1; b = (f2 - f1) / (x2 - x1) and c = [(f3 - f1) / (x3 - x1) – b] / (x3 – x2)
Minimum of qa(x) is at x = {(x1 + x2) / 2} – {b / (2 c)}

First Iteration/Approximation

x1 = 1.0, f1 = 51.0; x2 = 2.0, f2 = 29.0; x3 = 4.0, f3 = 28.5

xm = 4.0, fm = 28.5; a = 51.0, b = –22.0, c = 7.25

qa1 = 51.0 – 22.0 (x – 1.0) + 7.25 (x – 1.0) (x – 2.0)

6
Polynomial Approximation Methods
x = (x1 + x2) / 2 – b / (2 c) = 3.017, f = 25.68

Check for convergence.

Second Iteration/Approximation

x1 = 2.0, f1 = 29.0; x2 = 3.017, f2 = 25.68;


x3 = 4.0, f3 = 28.5; xm = 3.017, fm = 25.68;

a = 29.0, b = –3.265, c = 3.067

qa2 = 29.0 – 3.265(x – 2.0) + 3.067(x – 2.0)(x – 3.017)


x = 3.041, f = 25.69
Converged?

Вам также может понравиться