Modeling, Simulation and Optimisation

for Chemical Engineering

CH3133 – CC101, CC102, CC103

K2019 English Program
Ho Chi Minh City University of technology
Lecturer: Associate Professor Dr. Shuhaimi bin Mahadzir
PART III: Process Optimisation

Introduction to Process Optimization

Introduction to Process Optimization
The principle of optimization
Decision making:
“De doubus malis, minus est semper aligendum”
Translation:
Of two evils, always choose the lesser

Optimization is selecting the best among the entire set by efficient

quantitative methods.
Introduction to Process Optimization
The principle of optimization
Optimization in chemical engineering
• determining which chemicals and in what quantities to add to a mixture being
made so that it will meet specifications on its composition
• allocating available funds among various process improvement projects
• deciding which route to take to deliver products to several locations

Decision making always involves making a choice between various

possible alternatives
Introduction to Process Optimization
The principle of optimization
Essential features of optimization problems
• At least one objective function to be optimized (e.g. cost, profit, etc.)
• Equality constraints (e.g. process models)
• Inequality constraints (e.g. lower and upper bounds of process variables)

A feasible solution is a set of variables that satisfy the constraints.

An optimal solution is a set of values of the feasible solution that also

satisfy the minimum or maximum value of the objective function.
Introduction to Process Optimization
The principle of optimization
Example 1: Optimum reactor selection

Alternative 1: A simple low conversion reactor,

→ low reactor cost but increasing separation costs.

Alternative 2: A complex high conversion reactor,

→ high reactor cost but lowering separation costs.
Introduction to Process Optimization
The principle of optimization
Example 1: Optimum reactor selection

• X ↓ reactor cost ↓ separation cost ↑

• X ↑ reactor cost ↑ separation cost ↓

• Minimum total cost exists at Xopt.

Introduction to Process Optimization
The principle of optimization
Design Variables

• Design variable is always involved with a trade-off.

• Changing the value of a variable may bring benefits in one part of

the design but has detrimental effects in other parts.

• There must be an optimum value for the design variable.

Introduction to Process Optimization
The principle of optimization
Types of variables
• Continuous - can take any value in a region,
e.g. temperature, pressure, flowrate, …

• Discrete (e.g. integer) - can only take discrete values,

e.g. no. of reactors, no. of columns in a sequence, no. of stages, ...

• Binary variables - can only take 0 or 1,

e.g. existence or non-existence of equipment in a design, ...
Introduction to Process Optimization
The principle of optimization
Objective function
• Objective function (performance index) is expressed in terms of
design variables.

• The goal of optimization is to find the values of the variables that

yield the optimal value of the objective function.
Introduction to Process Optimization
The principle of optimization
Process model and constraints
• Constraints arise from kinetics, material and energy balances.
Physical bounds on variables, empirical relations, physical laws, …

• Constraints are expressed in form of equalities or inequalities.

• The whole set of equalities and inequalities plus the objective

function comprises the process model, which describes the inter-
relationships of design variables.
Model Assessments
Example 2:
Optimization of a heat exchanger
Given: Cps, Cpw = Specific heat capacities
C = unit cost of water
U = overall heat transfer coefficient
R = annualization factor
Exchanger capital cost ($) = a×A0.84
Subscripts: w = tube; s = shell side
Superscripts : s = supply temperature; t = target temperature

Objective: Design a minimum cost heat exchanger

(Assume DTmin = 10oC)
Model Assessments
Example 2:
Optimization of a heat exchanger
Variables: Q, Fw, Twt, A
Energy balance:

Feasibility: A, Q, Fw , Twt ≥ 0; Ts -Twt ≥ 10

Objective function: f = R(a×A0.84) + C×Fw
Model Assessments
Example 2:
Optimization of a heat exchanger
The mathematical programming model for optimization:
Minimize f = R(a×A0.84) + C×Fw  Objective function
Subject to: h1:
h2:
Equality constraints
h3:

g1: A, Q, Fw , Twt ≥ 0;
Inequality constraints
g2: Ts -Twt ≥ 10
Introduction to Process Optimization
The principle of optimization
Mathematical Programming Model
The general mathematical programming model for optimization can be
expressed in the following format:
minimise, or f(x) - objective function
maximise
subject to: h(x) = 0 - equality constraints
g(x) ≥ 0 - inequality constraints
where x is assumed as a set of continuous variables.
Introduction to Process Optimization
The principle of optimization
Mathematical Programming Model
• If f(x), h(x) and g(x) are all linear, then this leads to a linear
programming problem (LP).

• If objective function or at least one of the constraints is non-linear, it

becomes a non-linear programming problem (NLP).

• If some variables are integer, then we have either mixed integer

programming problem (MIP).
Introduction to Process Optimization
The principle of optimization
Degrees of Freedom (DOF)
Minimise f = x2 - x1 + 2x32 - 4x4x5
Subject to:
x3 - 2x1 + x2 = 2
x4 + x1 - 3x2 = 2
x5 + x 1 + x 2 = 4

The number of equations (NE) is 3 and

the number of variables (NV) is 5. DOF = 2
Introduction to Process Optimization
The principle of optimization
Exercise:
Minimize f = x2 - x1 + 2x32 - 4x4x5
Subject to:
x3 - 2x1 + x2 = 2
x4 + x1 - 3x2 = 2
x5 + x 1 + x 2 = 4
Analyse the DOF if you add:
• one more independent equation without introducing new variables
• two more independent equations without introducing new variables
• more than two equations, without adding new variables
Introduction to Process Optimization
The principle of optimization
Continuous and Discontinuous Functions
Introduction to Process Optimization
The principle of optimization
Continuous and Discontinuous Functions

A function of a single variable x is continuous at point x0 if,

a) f(x0) exists

b) lim f(x) exists

x → xo

c) lim f(x) = f(x0)

x → xo
Introduction to Process Optimization
The principle of optimization
Continuous and Discontinuous Functions
In optimisation it is more convenient to work with continuous functions
than discontinuous functions, e.g. search for a minimum
Introduction to Process Optimization
The principle of optimization
Unimodal and Multimodal functions
Introduction to Process Optimization
Linear Programming
General features of Linear Programming Problem
• LP is the most widely used optimisation technique in chemical industry
• LP implies that objective function and all constraints are linear
• LP problems are convex programming, where the objective function is
convex (linear) and the constraints form a convex region or solution
space Y an optimal solution is global optimum
Introduction to Process Optimization
Linear Programming
Formulation of Linear Programming Optimization Model
The basic steps involved in formulating an LP model are
1) Identify decision variables
2) Express constraints as linear equalities or inequalities
3) Write the objective function as a linear function
Model Assessments
Example 3:
An refinery processes 2 types of crude oil and produces 4 types of
products. The objective is to maximize the profit of the refinery
Model Assessments
Example 3:
Steps of problem formulation
Step 1 - Define variables:
x1 = bbl/day of crude 1 consumed
x2 = bbl/day of crude 2 consumed
x3 = bbl/day of gasoline produced
x4 = bbl/day of kerosene produced
x5 = bbl/day of fuel oil produced
x6 = bbl/day of residual produced
Model Assessments
Example 3:
Steps of problem formulation
Step 2 - Formulate
a) Equations:
Gasoline 0.8x1 + 0.44x2 = x3
Kerosene 0.05x1 + 0.1x2 = x4 c) Although not explicitly
Fuel oil 0.1x1 + 0.36x2 = x5 stated, all variables are
Residual 0.05x1 + 0.1x2 = x6 nonnegative
x1, x2, x3, x4, x5, x6 > 0
b) Inequalities:
Gasoline x3 ≤ 24000 ; or 0.8x1 + 0.44x2 ≤ 24000
Kerosene x4 ≤ 2000 ; or 0.05x1 + 0.1x2 ≤ 2000
Fuel oil x5 ≤ 6000 ; or 0.1x1 + 0.36x2 ≤ 6000
Model Assessments
Example 3:
Steps of problem formulation
Step 3 - Formulate objective function
Maximise f(x) = profit = income - raw material cost - processing cost
• Income: 36x3 + 24x4 + 21x5 + 10x6 ($/day)
• Raw material cost: 24x1 + 15x2 ($/day)
• Processing cost: 0.5x1 + x2 ($/day)
→ Max f = 36x3 + 24x4 + 21x5 + 10x6 - 24.5x1 - 16x2
The dimensionality of the problem can be reduced:
→ f(x) = 8.1x1 + 10.8x2
Model Assessments
Example 4:
A company produces chemical products in 1000 kg batches. A solvent
product is a mixture of two chemicals,A and B. The cost per kg of the two
chemicals are: Chemicals Cost/kg
A $3
B $5
Each batch has the following requirements:
a) at least 500 kg of chemical A
b) at least 200 kg of chemical B
The ratio of A:B must be at least 2:1.
The objective of the optimization is to determine the optimal mixture of
A and B that will minimize cost.
Model Assessments
Example 4:
Decision variables:
x1 = kg A to be used
x2 = kg B to be used

Objective function:
Min f = 3x1 + 5x2, where f = cost per 1000 kg batch.

Equations: x1 + x2 = 1000
Inequalities
x1 ≥ 500 x1/x2 ≥ 2
x2 ≥ 200 x1 ≥ 0, x2 ≥ 0
Introduction to Process Optimization
Linear Programming
Exercise:
A farmer is preparing to plant a crop and needs to fertilize a field. There are 2 brands
of fertilizer to choose from, Super-gro and Crop-quick.
Each brands yield a specific amount of nitrogen and phosphate per bag as follows:
Chemical Contribution
Nitrogen Phosphate
(kg/bag) (kg/bag) Over to
Super-gro 2 4 you
Crop-quick 4 3
The farmer requires at least 16 kg of N2 and 24 kg of phosphate. Super-gro cost
$6/bag and Crop-quick costs RM 3/bag. The farmer wants to know how many
bags of each brand to purchase in order to minimize the total cost of fertilizing.
Introduction to Process Optimization
Linear Programming
Solutions of LP
Consider the problem:
Max f(x) = 8.1x1 + 10.8x2
Subject to: 0.8x1 + 0.44x2 ≤ 24000
0.05x1 + 0.1x2 ≤ 2000
0.1x1 + 0.36x2 ≤ 6000
{x1, x2} ≥ 0
The LP model can be represented graphically.
Introduction to Process Optimization
Linear Programming
Solutions of LP
Feasible region: 0.8x1 + 0.44x2 ≤ 24000 (A)
0.05x1 + 0.1x2 ≤ 2000 (B)
0.1x1 + 0.36x2 ≤ 6000 (C)
x1≥ 0
x2≥ 0
Introduction to Process Optimization
Linear Programming
Solutions of LP
Contour plot:

Graphical solution to the problem

indicates that optimum value of the
profit occurs at:
x1 = 26000 bbl/day
x2 = 7000 bbl/day
f(x*) = 286200 $/day
Introduction to Process Optimization
Linear Programming
Exercise:
For the problem of crop fertilizing earlier, use the graphical method to
determine:
1. Optimal no. of bags for Super-gro fertilizer
2. Optimal no. of bags for Crop-quick fertilizer
3. The minimum total cost of fertilizing