Check certain relations before running Parameter Correlation

MarkAMarkA Member Posts: 5

Hello everyone,

In my current simulations I am modeling the flow through a nozzle, which connects a micro-channel to a fluid chamber. The model is in 2D and a pressure BC (pressure-outlet, 0 Pa static) is applied to the inlet of the micro-channel and another pressure BC (pressure-inlet, 1 kPa static) is applied to the top wall of the fluid chamber. A parametric study needs to be conducted to determine the influence of various geometric parameters. These parameters are shown in the figure below. Before conducting a DOE, I want to perform a Parameters Correlation study to find the most important parameters.

The nozzle is completely defined by 3 parameters: the nozzle throat width V11, the nozzle angle A13 and the nozzle length L12. The nozzle opening towards the chamber is reference dimension (L14) and is driven by these parameters according to (L14) = 2*(L12*tan(A13) + 1/2*V11). The fluid chamber height is also a parameter in the parametric study, indicated by H in the figure above. The vertical position of the nozzle w.r.t. the chamber is given by L33. The height of the microchannel is equal to (L14) at all times by means of a constraint.

I am able to set a lower and upper bound for the input parameters, but in order for the system to be valid, some relations must be satisfied. For example, the chamber height must at all time be larger than the nozzle opening towards the chamber, i.e H > (L14). Another example is that L33 > 1/2*(L14) such that the lowest point of the nozzle opening towards the chamber is not below the bottom of the chamber.

One option is to specify these relations as output parameters by means of an "if else" statement (1 if H > (L14) else 0 (H and (L14) replaced by their respective P#, (L14) according to expression stated above), and deleting all design points with 0), which means that I know whether the design point was valid after the simulation. However, ideally, these relations are checked before running the simulation of the design point in order to reduce the computational time of the complete parametric study.

The second option, which is my current workaround, is to

  1. Let Ansys generate a great amount of design points (way more design points than desired/required)
  2. Export the table with generated design points to a .csv file
  3. Open the .csv file in Matlab
  4. Check the relations and remove design points that fail any of the relations
  5. Save a as a new .csv file
  6. Import the new table into Ansys.

However, this method may cause certain input parameters to be unequally explored and therefore an unaccurate correlation matrix may result.

Is there a way to check these relations before starting the simulations, i.e. to not have them as output parameters?


Best regards.

Comments

  • peteroznewmanpeteroznewman Posts: 11,380Member
    edited December 2020

    @MarkA

    I suggest you rebuild the sketch to reduce interactions that can cause invalid domains without extensive constraints to keep it valid.

    An example of that would be to have 8 dimensions as the independent inputs as shown in the sketch below. The only constraint needed to assure a valid domain is to require all the inputs to be positive. The sketch is completely defined by various constraints such as symmetry about the X axis. You seemed to want a non-symmetrical volume on the right so there are two values to define how close the wall of the domain comes to the outlet of the nozzle. I choose the angle A3 to be an input. I could have used the height of the nozzle outlet as an input instead and measured the angle as an output parameter. That would have made it easier to calculate H. I did that in the second Archive. Two workbench archives are attached if you want a closer look.


  • MarkAMarkA Posts: 12Member

    @peteroznewman

    Thanks for this simple suggestion. It simplifies my problem tremendously.

    Cheers.

  • MarkAMarkA Posts: 12Member

    Dear @peteroznewman


    I have a follow-up question.

    In the picture below, a diffuser is added on the other side of the chamber. This diffuser is parametrized similar to the nozzle per your suggestion. However, how can I ensure that the position of the diffuser stays within the limit of the chamber height, i.e. such that V5+L2+V6 >= V10+L16?

    Right now in my DOE I used the limits [0,05 0,2] for V5, V6 and V10, and [0.2 1.0] for L2.

Sign In or Register to comment.