Strange outputs for fitness values


I am currently working on a parametric bridge model which I want to optimize with Wallacei. For the structural analysis I am using Karamba3D. As objective values I inserted:

  • the mass of the structure (kg)
  • the max. deformation (cm)
  • the natural frequency of the structure (Hz)
  • the utilization of elements (which shall be between 0.50 and 0.80)
    The genes are number sliders and gene pools representing the geometry boundaries of the bridge.

When I run the algorithm, the output of fitness values for each solution don’t seem to be realistic (especially for the deformation and the utilization which are way too high, see figure).
I expect the deformation values (FV 2) to be about 0…30cm at maximum. The FV 4 should be, as said, between 0.5 and 0.8.

If I reinstate some solutions of the last generation and compare the outputted values without the optimization, the calculation with Karamba3D seems to be working correctly.
So I expect the error to be at what I fed into Wallacei.

Dies anybody know what could be the reason for this?
Are the fitness objective on the output size somehow added together, which would explain the high values?

Since this is part of my master thesis I wish not to upload the .gh-file. I’ll try to explain it with screenshots.
Thank you so much in advance!

Best wishes,

Well, I think I found the solution for this in a recent post.
Looks like there has been an issue related to the region settings in windows. As to be seen on the screenshots above, the values after the optimization included a comma. That seemed odd to me.
I changed the the ‘decimal symbol’ from comma(,) to dot(.) in the windows settings.

Now the output values are displayed as expected.


I can also refer to this post:

1 Like