Modeling, Simulation and Optimization for

Chemical Engineering
Part 3: Optimization for Chemical Engineering

Lecture 12: Single Variable Optimization and Methods for Optimization

Ho Chi Minh City University of Technology

Contents
• Multi-objective  Addresses Real-World Complexity
Optimization
• Methods for
Multi-Objective
 Enhances Decision-Making
Optimization
• Unconstrained
Multivariable
 Balances Trade-Offs
Optimization
• Methods for
 Promotes Resource Efficiency
Wide Applicability
Unconstrained
Optimization 
 Drives Sustainable Solutions

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective • Some needs to increase
Optimization
• Methods for
Multi-Objective
Optimization
• Some needs to decrease
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Paper Titles Authors
• Multi-objective
Optimization
• Methods for
Halim,I., Srinvasen,R.
Integrated decision support system for waste
Multi-Objective National University Singapore,
minimization analysis in chemical processes
Optimization 2002
• Unconstrained
Multivariable Mauricio M. D. & Karen, A.H.
Evaluation of waste minimization alternatives
Optimization Oklahama state university, USA
• Methods for under uncertainty: a multi-objective approach
1999
Search
Direction Analyzing the interaction of design Motivation for
Luyben, M.L., Flounders, C.A.,
Multiobjective Optimization and control. A multi-
Computers and Chemikal
objective framework and application to binary
Engg., 18, 933 – 969, 1994
distillation synthesis
Cavin Laurent Swiss Federal
A systematic approach for multiobjective process
Institute of Technology Zurich.
design in multipurpose batch plants
2003

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective Paper Titles Authors
Optimization
• Methods for
Multi-Objective
Assessment of chemical risk and safety by the use
Garribba, Ovi GASP, Italy, 1980
Optimization of multi-attribute decision techniques
• Unconstrained
Multivariable
Busacca, Marseguerra, Zio
Optimization Multi-objective optimization by genetic algorithms:
• Methods for Dep. of Nuclear Eng.,
application to safety systems
Unconstrained CESNEF, Italy, 2001
Optimization

Kim, Raymond, Smith U.S.

Systematic procedure for designing processes with
Environmental Protection
multiple environmental objectives
Agency, 2005

Multi-objective optimization for safety-related Kim, Yeo, Moon Yonsei

decision making in chemical processes University Korea 2004

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
• Involve more than one objective function that are to be
Multi-Objective
Optimization
• Unconstrained
minimized or maximized.
Multivariable
Optimization
• Methods for • Answer is set of solutions that define the best tradeoff
Unconstrained
Optimization
between competing objectives.
• There is no unique solution to Multi-objective Optimization
problems.

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• In the single-objective optimization problem, the superiority of a
• Methods for
Multi-Objective
Optimization
solution over other solutions is easily determined by
• Unconstrained
Multivariable
Optimization
comparing their objective function values
• Methods for
Unconstrained
Optimization • In multi-objective optimization problem, the goodness of a
solution is determined by the dominance

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• x1 dominates x2, if
• Methods for
Multi-Objective
Optimization
• Unconstrained
• Solution x1 is no worse than x2 in all objectives
Multivariable
Optimization
• Methods for • Solution x1 is strictly better than x2 in at least one objective
Unconstrained
Optimization

• x1 dominates x2 x2 is dominated by x1

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization • 1 Vs 2: 1 dominates 2
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable • 1 Vs 5: 5 dominates 1
Optimization
• Methods for
Unconstrained
Optimization • 1 Vs 4: Neither solution

dominates

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
Non-dominated solution set
• Methods for
Multi-Objective
Optimization
• Unconstrained
Given a set of solutions, the non-dominated solution set is a set
Multivariable
Optimization
of all the solutions that are not dominated by any member of the
• Methods for
Unconstrained
solution set
Optimization

• The non-dominated set of the entire feasible decision space is

called the Pareto-optimal set

• The boundary defined by the set of all point mapped from the
Pareto optimal set is called the Pareto optimal front
Lecture 13: Multi-Objective Optimization
Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Find set of solutions as close as
• Methods for
Multi-Objective
Optimization
• Unconstrained
possible to Pareto-optimal front
Multivariable
Optimization
• Methods for
Unconstrained
• To find a set of solutions as
Optimization

diverse as possible

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Two important methods are:
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
• Weighted sum method
Optimization

• ε-Constraint Method

Lecture 13: Multi-Objective Optimization

Contents Converts multiple objectives into a single objective by
• Multi-objective
Optimization assigning weights
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization wi Weights assigned to each objective function fi(x)
• Methods for
Unconstrained
Optimization Steps:
Assign weights to objectives based on importance.
Solve the optimization problem for the aggregated function.
Example:

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for Convex Non-Convex
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Advantages:
• Multi-objective
Optimization
• Methods for
Multi-Objective
• Easy to implement.
Optimization
• Unconstrained
Multivariable
• Reduces the problem to a single-objective optimization.
Optimization
• Methods for
Unconstrained
• Works well when objectives are not highly conflicting.
Optimization
Disadvantages:
• Sensitive to weight selection.
• May miss Pareto-optimal solutions in non-convex regions.
• Requires knowledge of trade-offs between objectives.
Lecture 13: Multi-Objective Optimization
Contents Example Problem:
• Multi-objective
Optimization
• Methods for
• Optimize a chemical process with:
Multi-Objective
Optimization
• Objective 1: Minimize energy cost.
• Unconstrained
Multivariable
• Objective 2: Minimize environmental impact.
Optimization
• Methods for
Unconstrained
Solution:
Optimization

Lecture 13: Multi-Objective Optimization

Contents Converts a multi-objective problem into a single-objective
• Multi-objective
Optimization problem by optimizing one objective while restricting others:
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization Steps:
1.Select one primary objective to optimize.
2.Set upper bounds (constraints) on other objectives.

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization • Keep f2 as an objective;
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
minimize f2(x)
Optimization
• Methods for
Unconstrained
Optimization • Treat f1 as a constraint;

f1(x) ≤ ε1

Lecture 13: Multi-Objective Optimization

Contents Advantages:
• Multi-objective
Optimization
• Methods for • Generates Pareto-optimal solutions systematically.
Multi-Objective
Optimization
• Unconstrained • Effective for non-convex problems.
Multivariable
Optimization
• Methods for Disadvantages:
Unconstrained
Optimization
• Requires careful selection of ϵ-values.
• Computationally expensive for problems with many
objectives.
• May miss solutions if ϵ-values are poorly chosen.
Lecture 13: Multi-Objective Optimization
Contents Example Problem:
• Multi-objective
Optimization
• Methods for • Optimize a distillation column operation:
Multi-Objective
Optimization
• Unconstrained Objective 1: Minimize energy usage.
Multivariable
Optimization
• Methods for Objective 2: Maximize purity.
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Definition:
• Multi-objective
Optimization Finding the optimal value (minimum/maximum) of a function with
• Methods for
Multi-Objective multiple variables, without constraints on the variables.
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained Key Points:
Optimization
• No restrictions (e.g., bounds or equality constraints) are
imposed on xi
• Common in engineering, machine learning, and physics.
Examples:
• Minimize energy consumption in a process.
• Maximize efficiency in a chemical plant.
Lecture 13: Multi-Objective Optimization
Contents Gradient ( f)
• Multi-objective
Optimization
Direction of the steepest ascent/descent.
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for Hessian Matrix (H(f))
Unconstrained
Optimization Matrix of second-order partial derivatives, used to evaluate
curvature.

Lecture 13: Multi-Objective Optimization

Contents Direct Methods
• Multi-objective
Optimization
• Methods for
Do not require derivative information.
Multi-Objective
Optimization
• Unconstrained
Use only function evaluations to decide the search direction.
Multivariable
Optimization Examples: Random Search, Grid Search, Simplex Method.
• Methods for
Unconstrained Indirect Methods
Optimization
Require gradient or Hessian information.
Use this information to compute the search direction.
Examples: Gradient Descent, Newton’s Method.

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Random search method
• Multi-objective
Optimization
• Methods for
A random search method simply selects a starting vector xo, evaluates
Multi-Objective
Optimization
• Unconstrained f(x) at xo , and then randomly selects another vector x1 and evaluates
Multivariable
Optimization
• Methods for f(x) at x1. In effect, both a search direction and step length are
Unconstrained
Optimization
chosen simultaneously. After one or more stages, the value of f (xk) is

compared with the best previous value of f(x) from among the previous

stages, and the decision is made to continue or terminate the procedure.

Lecture 13: Multi-Objective Optimization

Contents
• Multi-objective Grid Search Method
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Univariate Search Method
• Multi-objective
Optimization
• Methods for
• This method searches one dimension at a time thus optimizing only a
Multi-Objective single variable per iteration.
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents Simplex Search Method
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable
Optimization
• Methods for
Unconstrained
Optimization

Lecture 13: Multi-Objective Optimization

Contents A good search direction reduces f(x) such that:
• Multi-objective
Optimization
• Methods for
Multi-Objective
Optimization
• Unconstrained
Multivariable Such a direction is called a descent direction.
Optimization
• Methods for
Unconstrained
Optimization

Methods
• Steepest Decent Method
• conjugate gradient method

Lecture 13: Multi-Objective Optimization