doe Package

What is the doe package?

The Design of Experiment package supports design space exploration functionalities for Multibody systems. It aims to simplify the process of parameter optimization and sensitivity analysis in simulation models. Multiple experiment design types are supported, ranging from factorial to randomized designs. User-defined design is also available.

This package introduces the DesignExperiment class, which is responsible for the setup and execution of the DOE. After choosing the best design type for your application, you can specify the design parameters (see Dv) and response variable(s) (see Response Module) of your model to be included in your DOE. If no responses are given to the DesignExperiment class, the model will be examined for pre-existing responses and automatically added to the class. On setup, a stack of different simulation cases is created. Those cases are scattered throughout the design space based on the design type. On run, all the different simulation cases are executed and the DesignExperiment is populated with the simulation and the response results.

Parallel execution is supported, leveraging the multiprocessing/threading libraries of python. This can dramatically reduce simulation times, especially in extensive design exploration, with multiple parameters and/or run cases. Parallel execution is particularly effective on workstations or a high performance cluster, such as provided by Altair Appliance or AltairOne.

Use the ReportDOE class to assess the DOE’s quality. It generates a Jupyter Notebook file (.ipynb) containing the DOE results. When you run the notebook, it produces graphics and plots to show the sensitivity of the response(s) in respect to the design variable(s).

DesignExperiment Class

Establishes a set of simulations to investigate the influence of design variables on responses.

Response Module

Module that entails a library of classes currently available for creating commonly used responses.

ReportDOE Class

Creates a Jupyter Notebook (.ipynb) document that provides a summary of the executed DOE.

DOE Example

Provides an example on how to define and execute a DOE.