0 to optimization in 10 steps

0 to optimization in 10 steps

This guide serves as a medium-length approach to developing an understanding of how to perform multidisciplinary design optimization (MDO) assuming no prior knowledge of MDO. We assume that you have college-level training in engineering concepts, including differential equations, Python programming, and basic numerical methods. Additionally, most all of the examples use OpenMDAO, so having experience with it or going through the documentation is beneficial.

Also, step 10 could arguably come first as it might inform how you set up the filesystem for these models and optimizations.

  1. Building uncoupled feed forward models

    1. Understanding XDSM diagrams

    2. Using groups to organize models

    3. Connecting vs promoting variables

    4. Using the N2 diagram

  2. Problem formulation

    1. Basic optimization problem formulation

    2. Multiobjective optimization

  3. Simple gradient-free opt

    1. When to use gradient-free optimizers

    2. Types of gradient-free methods

  4. Get partial derivatives

    1. A brief introduction to derivatives

    2. Common ways to compute derivatives

  5. Gradient-based opt

    1. [[Why to use gradient-based optimizers]]

    2. [[Types of gradient-based optimizers]]

  6. Building coupled models

    1. Explicit vs implicit systems

    2. [[Fixed point iterators vs Newton]]

  7. [[Gradient-free multidisciplinary optimization]]

  8. Total derivatives for coupled models

    1. Nonlinear and linear systems and solvers

    2. Total vs partial derivatives

  9. Gradient-based multidisciplinary design optimization

  10. [[Relevant software engineering topics]]

    1. [[Git]]

    2. [[Folder and package structure]]

    3. [[Tests]]

    4. [[Continuous integration]]

previous

Guided lessons

next

What you need for gradient-free optimization