Parameter driven real-life design Process through star ... · PDF file©p.èn. lab & Electrolux 2011_ ParamEtEr drivEn rEal-lifE dEsign ProcEss through star-ccm+ ... • Parameter

©p.èn. lab & Electrolux 2011_ ParamEtEr drivEn rEal-lifE dEsign ProcEss through star-ccm+ automation

Parameter driven real-life design Process through star-ccm+ automationMauro Poian p.èn. labCedric Catalogne Electrolux

| 01©p.èn. lab & Electrolux 2011_ ParamEtEr drivEn rEal-lifE dEsign ProcEss through star-ccm+ automation

Everything inside a CFD model can be considered a parameter.•

With the introduction of the 3D-CAD module starccm+ extends its capabilities into the parametric field. •

The well-established API system provides a robust framework for automation.•

Why Parameter driven design?


| 02

traditional Parameter driven design

Macro has to be manually created and modified.•

Parameter study are not linked to any specific D.o.E. strategy.•

Simulation structure and maintenace has to be implemented.•

Output has to be manually collected and organized.•


| 03

What design manager does

Allows to prepare a list of design (sequence of parameter combination) importing it from external files or •creating it by using D.o.E. algorithms.

Update the model accordingly to the new parameters.•

Stores the .sim in a well-organized directory structure.•

Collect and export results in a structured format.•


| 04

hoW the design manager Works

Design Manager

Starccm+

Reports PlotsModel parameters

Geometrical parameters Analyses monitoring

Java Macro


| 05

available features

Import from external ASCII files.•

D.o.E. techniques: • - Random - Factorial - Geometrical

Automatic model regeneration and analysis.•

Data mining.•

Data Visualization and export in ASCII format. •


| 06

design manager: setuP


| 07

design manager: actions


| 08

industrial examPle of Parameter driven desing oPtimization

Washing Machines are getting smarter and smarter •thanks to the intelligence provided by electronic controllers.

These devices are placed in a very compact package •subjected to extremely severe thermal condition: CFD analysis is mondatory to virtually test the performance of the system before testing and verification.

This integrated control requires cooling power in order •to discharge the heat release by the IGBT module. This cooling is ensured by natural convection on the heat sink (Aluminium alloy type Pyral or similar) which design needs to be optimized.


| 09

model set-uP: initial investigation The overall system was first the object of an analysis through an “extensive” CFD model.

It allowed to understand the critical point as well as identified model simplification before running the actual optimization.


| 10

model set-uP: initial investigation A simplified model, focusing on the heat sink, was set up in order to evaluate the temperature of the aluminum heat sink.

This simplified “stand-alone” model was integrally managed through star-ccm+. dEsign ParamEtErs

Frame thickness. •Number of wings. •Wings height.•Wings thickness. •

Pdiss = 14 W

tamb = 60°c


| 11

design strategy A DoE based optimization and response surface methodology was adopted:

A face centered cubic design composed of a full or fractional factorial design and star points placed on the faces of the sides is chosen and repeated for each wing number.

1 - A Design of experiment (DoE) is defined.2 - CFD simulation is run for each of the DoE configurations. 3 - The resulting data set is interpolated with a response surface (RSM). 4 - The optimum(s) are extracted from this RSM considering the geometrical constraints on the heat sink geometry. 5 - Verification of the optimal solutions (simplified model). 6 - Verification using the full model.


| 12

automatic WorkfloW

An automatic process was defined to speed the turn round evaluation time per design.•

The parametric geometry is directly created in the CAD modeler embedded in Star-ccm+ in order to create •a continuous process flow.

Each design has been created using the D.o.E. algorithms available in the Design Manager.•

Each design evaluation is automatically computed by the Design Manager (Parameter update, mesh/solver, •output data mining).

The full process time was of 86 hours for the 60 cases. •

1 min.

GEOMETRY UPDATE MESH & MODEL SET-UP SOLVING POST-PROCESSING

15 min. 200 min. 1 min.


| 13

design manager automation

GEOMETRY UPDATE

MESH & MODEL SET-UP

SOLVER

POST-PROCESS


| 14

scatter visualization


| 15

rsm interPolationWeight

Temperature


| 16

rsm full factorial oPtimization


| 17

rsm interPolation: the feasible region


| 18

oPtimal configurations

Optimum2.

Weight

(-27%)

T al

OK /+3%

Optimum3.

Weight

(-18%)

T al

OK /+0%

NominalWeight

Target

T al

Target

Optimum1.

Weight

(-34%)

T al

OK /+7%


| 19

conclusion

Optimum 2 was selected as a trade-off choice and •adapted to the geometrical and process constraint.

Reviewed and final heat sink design is 21% less •costly and ensure a maximum temperature under the target limit.


| 20

What’s next… Automation Framework.

D.o.E. techniques.

Surrogate Models.•

Full Automated Optimization.•


Parameter driven real-life design Process through star-ccm+ automationMauro Poian p.èn. labCedric Catalogne Electrolux

Thank you.

Documents

Parameter driven real-life design Process through star ... · PDF file©p.èn. lab & Electrolux 2011_ ParamEtEr drivEn rEal-lifE dEsign ProcEss through star-ccm+ ... • Parameter