Opti­miza­tion of a Cat­alytic Converter

The design of engine com­po­nents for cars is a tricky task, because there are so many con­straints involved. An example is the duct that is located right before the cat­alytic con­verter. This com­po­nent is often bent quite heavily due to space con­straints. As a result, this makes it a tough chal­lenge to design it in a way that the flow dis­tri­b­u­tion is suf­fi­ciently uniform. In other words, bad flow char­ac­ter­is­tics in the cat­alytic con­verter lead to poor per­for­mance, and possibly higher emissions.

In one of our customer projects we had worked on the opti­miza­tion of such a duct, where three tools were involved:

• CAESES® for geometry modeling and the auto­mated optimization
• GridPro for meshing the gen­er­ated design candidates
• OpenFOAM for CFD analysis

In the fol­low­ing sections, we briefly outline the workflow of this design task.

Geometry for Robust Duct Variation

CAESES® was used to create an intel­li­gent para­met­ric model of the duct, which was varied with a small set of suitable para­me­ters. We set up a meta surface which basi­cally swept a variable ellipse section. This means that the shape of the ellipse section along the path was varied by using function graphs, plus the sweep path was varied by design vari­ables, too.

We prepared the baseline design such that a closed geometry could be gen­er­ated for all sub­se­quent variants, includ­ing inlet and outlet bound­aries for which we used a coloring mech­a­nism. In CAESES®, the bound­aries can be named by setting custom color names and by assign­ing these colors to the surface patches. The fol­low­ing picture shows a typical colored STL output:

Setting up a closed geometry needs to be done only once, and then a user can recycle all settings during a design study or opti­miza­tion. The next ani­ma­tion shows auto­mated shape changes of the duct geometry: For each design a closed STL file is auto­mat­i­cally gen­er­ated and later on trans­ferred to the meshing tool. Note that in the ani­ma­tion below we show very strong changes, just for demon­stra­tion purposes. Usually, the geometry changes are much smaller when starting from a given baseline design.

Meshing and Flow Analysis

One of the key benefits of using GridPro is that a flow-aligned struc­tured mesh can be gen­er­ated quite easily, and it also adapts auto­mat­i­cally to a change in the geometry. The output mesh can be directly con­verted to the OpenFOAM format, and all of these GridPro commands are simply written into a shell script so that it can be auto­mated right away.

For the OpenFOAM CFD setup, we used a com­press­ible gas and involved a porous media for the cat­alytic con­verter. The control files of OpenFOAM were set up accord­ing to the customer require­ments (mass flow, tem­per­a­ture etc.). The OpenFOAM exe­cu­tion was again written into a shell script so that every­thing was ready for batch-mode automation.

Automa­tion and Optimization

Since the meshing and CFD process is con­trolled by shell scripts and the OpenFOAM control files, we could create a software con­nec­tor in CAESES®, to inte­grate and automate the entire process through the CAESES® user inter­face. The control files (dicts etc.) were loaded into CAESES®, and para­me­ter­ized for direct control of the settings. For an easier assess­ment, we also visu­al­ized the 3D flow data in the user inter­face of CAESES®, using the VTK output data from OpenFOAM.

The design viewer in CAESES® was also used for side-by-side com­par­isons of the gen­er­ated design can­di­dates. This usually includes geometry screen­shots and the 3D geometry itself, but also typical CFD results such as con­ver­gence plots for each design. With such an overview and the data from the result table, it is easy to finally pick the best can­di­date with maximum flow uniformity.

Results

The pressure loss was also taken into account in this project, at least up to a certain extent for finding a good com­pro­mise between flow uni­for­mity and pressure loss. As a result, we gained about 2.1% improve­ment for the uni­for­mity, while further reducing the pressure loss by 5.5%.