Chemical Processes Optimisation

2023/24
Introduction to Optimization
 Optimization Problem

Problem which translates a decision choice representable

in a mathematical form:

• At least, 1 objective function to be optimized (profit, cost, …) – f(X)

• Restrictions of the problem (equations and/or inequations) – gm(X)

2
 Optimization Problem

Limited
Resources

Formulation
Process
Limitations OBJECTIVE MODEL

Optimization
Technique

Pos-Optimization Optimal Solution

3
Optimization Problem Classification
Class Designation Objective Restrictions Variables
Function
Linear
LP Linear Linear Continuous
Programming
Non-Linear
NLP Non-linear Non-linear Continuous
Programming
Quadratic
QP Quadratic Linear Continuous
Programming
Integer
IP Linear Linear Discrete
Programming
Mixed Integer
MILP Linear Linear Linear Mixed
Programming
Mixed Integer
MINLP Non-Linear Non-linear Non-linear Mixed
Programming

4
Solution Methods

Analytical Methods Numerical Methods

f unimodal in the search
f and its derivatives
domain (but f and its
continuous in the search
derivatives can be
domain
discontinuous)
All the optimals are detected Only one optimal is detected
in the search domain in each search

5
 One-dimensional Functions

Conditions for the existence of a optimal (non-restrained

functions):
∗

Maximum
∗ ∗

Minimum
∗ ∗

∗ ∗
Inflexion Point:

Generally, n = 1

6
Concave and Convex Functions

Concave function (one-dimensional case)

• For any set of two values xa and xb, and a scalar 0≤ ≤ 1:

Convex function

Strictly Convex function

Strictly Concave function

7
Concave and Convex Functions

Convex function
(one-dimensional case)

Concave function
(one-dimensional case)

8
 Multidimensional Functions

Gradient Vector

Hessian Matrix
(Square and symmetric)

9
Multidimensional Functions

Eigenvalues Leading Principal

f(X) H(X)
H(X) Minors H(X)
Strictly Positive All diagonal values >0
>0
Convex Defined 1>0, 2>0, …
Positive All diagonal values ≥0
Convex ≥0
Semidefined 1≥0, 2≥0, …
Negative All diagonal values ≤0
Concave ≤0
Semidefined 1≤0, 2≥0, …
Strictly Negative All diagonal values <0
<0
Concave Defined 1<0, 2>0, …
Not
Concave Undefined Any other condition
or Convex

10
Necessary Condition

Necessary condition for the existence of a extreme

(minimum or maximum) in a multidimensional function:

X* - Stationary Point

11
Sufficient Conditions

Analysis of the Hessian Matrix type

Around X*,
H(X) XT.H(X*).X
f(X)-f(X*)
Positive >0
>0
Defined Minimum
Positive Analysis of higher order
≥0 derivatives
Semidefined
Negative Analysis of higher order
≤0 derivatives
Semidefined
Negative <0
<0
Defined Maximum
≤0 or ≥0
Undefined ≤0 or ≥0
Saddle Point

12
Summary

Necessary conditions for the existence of a extreme

(minimum or maximum) in a multidimensional function:

1. ∗

2. ∗ X*, Stationary Point

Sufficient conditions for the existence of a extreme in a

multidimensional function:

1. ∗ X*, Minimum

2. ∗ X*, Maximum

13