# Remove design points based on conditional relation for design optimization

Member Posts: 5

Hello all,

I am currently modeling a nozzle which allows fluid to flow from a micro-channel into a fluid chamber. There is a pressure inlet BC (1k Pa static pressure) at the left edge of the micro-channel and a pressure outlet (atmospheric pressure) on the top wall of the fluid chamber. I want to set up a parametric study to determine the optimal geometry of the nozzle. The system is in 2D.

The geometry is shown in the picture below.

The parameters in the parametric study fully define the nozzle shape. These parameters are the throad width V11, the nozzle angle A13 and the nozzle length L12. The nozzle opening width on the side of the chamber is added as a reference value (L14). Furthermore, the chamber height is set as a parameter (H, which has been added to the image due to the dimension being placed outside of the image window). L33 indicates the distance from the bottom wall of the chamber to the symmetry plane of the nozzle, and (L181) indicates the distance from the top wall of the chamber to the top of the nozzle opening (as a reference). Lastly, the height of the micro-channel is always equal to (L14).

In the Parameter Correlation/DOE part of my workbench project I am able to set lower and upper bounds on the throad width, angle and length of the nozzle, but in order to prevent design points with illegal geometries from being run, some constrains have to be fullfilled. For example, L14<H, and L33>(L14)/2.

How can I impose these relations in Ansys directly? Currently I have set the conditions as output parameters using if statements. For example, 1 if L14<H else 0 (with L14 and H being replaced by their parameter numbers). However, these output relations are calculated after the simulation has finished, which is undesirable since these relations need to be checked before the simulation is ran and new design points can be created.

My current workaround is to generate a few hundred design points, export the dp table to a .csv file, remove the illegal design points in Matlab and import the new table back into Ansys. This may result in a bad distribution of certain input parameters, and hence a better method is required.

Hopefully someone can help me find a solution to this problem.

Best regards.