Parametric modeling and optimization are very important topics these days for CFD engineers and researchers. Coupling OpenFOAM and Dakota can be a very powerful tool for your engineering workflow because it can be used very efficiently for design and optimization of your product.
OpenFOAM is one of the most popular and powerful CFD code available freely and widely used by several engineers and researchers. It is free, open source and does not require expensive licenses which may become prohibitive for very large scale simulation projects requiring hundreds of CPU cores.
Dakota is one of the best freely available optimization software packages boasting several optimization algorithms both open source and commercial. Dakota is a C++ optimization toolkit which can be used not only for optimization but also for uncertainty analysis, sensitivity analysis and calibration problems, with applications spanning defense programs for DOE and DOD, climate modeling, computational materials, nuclear power, renewable energy, and many others.
By combining OpenFOAM and Dakota, you can parameterize various practical engineering design problems and optimize your design. OpenFOAM can deal with almost all fluid dynamics problems you can face including incompressible and compressible flows, multiphase flows, non-Newtonian flows, conjugate heat transfer, cavitation, etc. On the other hand, Dakota is also a very capable code providing various optimization algorithms including gradient based methods, search based methods, genetic algorithms, hybrid methods, surrogate approaches, and so on.
So, if you want the optimal efficiency for your heatsink or if you are designing the bulbous bow of your container ship or if you just want to conduct a mesh convergence test, you can do it by coupling these two.
So, let’s cut to the chase and go into the main topic now. It’s HOW? To couple OpenFOAM with Dakota for parameter analysis, you need to parameterize the geometry. You can use open source CAD/CAM/CAE softwares with scripting ability such as Gmsh, Salome or FreeCAD. And you will need a driver code which will be run by Dakota for each optimization iteration. This is a shell script where you put all your workflow for one optimization iteration step.
So, the workflow is as follows.
- You create the base geometry in your CAD software and save the script. You also copy the script and name it script.template (of course you can name anything you want).
- Dakota generates the parameters to be used for the current step and writes to a file called parameters.in.
- dprepro python script provided by Dakota reads the parameters.in and script.template and rewrite your original CAD script by inserting parameters to template file. (I know, it’s frustratingly counterintuitive, huh. But that’s the way it’s supposed to run)
- CAD software reads the changed CAD script and generates the new STL file.
- snappyHexMesh reads the new STL and generates new OpenFOAMesh.
- Your solver (pimpleFoam, simpleFoam, interFoam, etc.) reads the mesh and calculates the flow.
- Probes/Forces/other function objects calculate your target variable and give to Dakota
- Dakota compares the current value and previous values and decides what value to use in next step.
For the case files, you can reference Dr. Holzmann’s tutorials here.
