| In the Computational Optimization course, the concepts of unconstrained and constrained optimizations are covered. The most important methods for systematic mathematical optimization are explained and applied using practical examples. In the first part of the course, local (gradient-based) optimization methods are covered in detail, while global optimization methods are covered in the second part of the course. Examples from structural mechanics will be discussed in the lectures and assignments. |
The course is a blend of mathematics, programming and engineering, and touches on a wide variety of Mechanical Engineering topics:
- Design optimization, design variables, objective function, constraints.
- Unconstrained optimization and line search.
- Constrained optimization and Lagrange multipliers.
- Local optimization methods for linear and non-linear problems.
- Calculation of gradients.
- Global optimization methods.
- Surrogate model based global optimization.
- Topology optimization.
- Practical approaches to optimization in engineering.
Why this course: In every engineering field, models are used to assess the performance of structures/products/systems. In order to improve the performance of a system, certain parameters of the design can usually be adapted. Computational optimization refers to the use of algorithms to automate the search for an optimal design. The adequate algorithm to be used depends on the characteristics of the problem, properties of the model and the number of parameters to be optimized.
Within this course, the following knowledge and skills will be developed:
- Formulation of a design assignment as an optimization problem;
- Classification of an optimization problem by its type of parameters, objective function and constraints;
- Choice of an appropriate mathematical solution algorithms for specific optimization problems;
- Usage of an optimization software toolbox to solve optimization problems.
Course highlights:
- Introduction to engineering optimization;
- Unconstrained optimization: Simplex method, Descent methods (Newton), Line search;
- Constrained optimization: Penalty methods, Lagrange multipliers, Quadratic and nonlinear programing; interior point method, augmented Lagrange method;
- Linear programming;
- Calculation of gradients: Finite differences, Algorithmic differentiation, Direct/Adjoint method;
- Global optimization: Pattern search, DIRECT method, Particle swarm, Genetic algorithms;
- Bayesian Optimization: Surrogate modelling, Kriging; Radial Basis Functions;
- Topology optimization: Solid Isotropic Material with Penalization.
For whom: Professionals with basic understanding and application of FEM and Matlab.
From whom: dr. J. Hazrati Marangalou
Practical information: This is a regular master course, in which students as well as professionals can participate. The course comprises optional lectures, of which handouts will be distributed. Practical assignments to apply the methods accompany the theoretical part. It is strongly recommended to follow the lectures. In an oral exam, the knowledge and skills will be evaluated.
Location: University of Twente, Enschede, NL
Duration: The course is scheduled annually from April till July. It requires 140 hours of study load.
Costs: € 2067,15
More information:
Content of the course: dr. J. Hazrati Marangalou, j.hazratimarangalou@utwente.nl
Registration: Registration form | Faculty of Engineering Technology (ET)