Optimization, Design of Experiments

main concepts

To set up a design exploration with the API, the user needs to know and understand some key concepts. Each concept is represented by an object in the ame_DE_apy module. The details of each object are as given in the following sections.

The DE Manager

the DE Manager

This is the topmost object in charge of handling all the studies for all the Simcenter Amesim models. This object is realized through the DEManager class.

The Model Studies Handler

the Model Studies Handler

The Studies handler is in charge of loading studies attached to a Simcenter Amesim model. It acts as a manipulator for the studies list related to the Simcenter Amesim model. It makes it possible to get the list of studies attached to the Simcenter Amesim model, to retrieve a given study, to create studies and to save the list of studies to the Simcenter Amesim DE model file. This object is realized through the ModelStudies class.

The Study

the Study

A study is the main entry point for a DE analysis. It gathers parameters (defined in the Simcenter Amesim export setup dialog), algorithms (analysis methods) and results of the DE analysis.

There are 3 possible types of study:

• DOE (Design Of Experiments),

• OPT (OPTimization),

• MC (Monte Carlo),

When you create a study, the study type (DOE, OPT or MC) is required so that the instantiated study has the expected type and can filter the algorithms accordingly. This object is realized through the Study class.

The Parameters

the Parameters

Parameters are defined from the Simcenter Amesim export setup dialog. They can either be input parameters or output parameters. This object is realized through the Parameter class.

Figure 7: Example of input parameters defined from Simcenter Amesim

Figure 8: Example of output parameters defined from Simcenter Amesim

The Algorithm

the Algorithm

Algorithm is the method used to compute a study. There are 10 possible algorithms. This object is realized through the Algorithm class.  Algorithms are applicable according to the study type, as shown in the table below:

ALGORITHM TYPE	VALID FOR STUDY NAMED
ParameterStudy_UniDirectional	DOE
ParameterStudy_FullCombinations	DOE
ParameterStudy_UserSets	DOE
FullFactorial	DOE
CentralComposite	DOE
NLPQL	OPT
GeneticAlgorithm	OPT
Optmized_LH	MC
Latin_Hypercube	MC
Random	MC

Note

Further details about each type of study and algorithm are given in the Design Exploration manual.

The Result

the Result

The concept of result makes it possible to retrieve all the results of a DE analysis which has been run. The result object encapsulates everything about the DE run, such as the number of simulations, parameter values according to the run, best parameter values and so on. Of course the content of the result depends on the parent study and specifically the parent study type. This object is realized through the Result class.

Error management

error management

An API throws exceptions when it is wrongly used or incorrect parameters are passed to it. The full list of exceptions that can be thrown by each API is detailed in the reference documentation.

Note

For more information on the concept of exceptions in Python please refer to the official Python documentation about exceptions which is available at