Guidance/suggestions with my results and problem setup.

Hi, I have managed to set up a problem with converging results and would like some guidance or suggestions to improve my set up. I would like to optimise the structure to reduce the overall mass. I have attached screenshots of the results I have obtained and it doesnt seem quite right to me. I was expecting a truss-like structure so I was wondering if anyone has any suggestions or recommendations. I have tried changing some of the MINDIM, MAXDIM, element size and shape but i have had no luck.

Just a summary of the problem setup.

- This is a guide vane which has a concentrated load of 11kN applied at the leading edge.

- SPCs on the bottom section

- Volume limit of 0.3

- Solid volume tetramesh

- Shell surface 2D tria mesh (1mm thickness)

- Objective set to minimise compliance

- average element size 3mm

- MINDIM 30 mm

- MAXDIM 80 mm

GuideVane3mm.hm

Find more posts tagged with

English

HyperMesh

OptiStruct

v2021.0

Accepted answers

All comments

PaulAltair

I think part of the issue you are seeing is that you are only optimising against a single loadcase (the leading edge point load) so the material distribution will be highly influenced by that 1 load.

A common approach is to run 3 or 4 different loadcases, for each, run a separate topology optimization and set a material volfrac constraint of e.g. 30% with target to minimise compliance.

Once you have the results (and compliances) for the individual loadcases, you can run another optimization where you run all the loadcases as part of the optimisation but set those compliances from the initial runs as constraints on those loadcases while minimizing volfrac as the objective (the resulting volfrac will be above 30%, but you then know each case is as good as if it were optimized for with 30% material)

D_21643

I think part of the issue you are seeing is that you are only optimising against a single loadcase (the leading edge point load) so the material distribution will be highly influenced by that 1 load.

A common approach is to run 3 or 4 different loadcases, for each, run a separate topology optimization and set a material volfrac constraint of e.g. 30% with target to minimise compliance.

Once you have the results (and compliances) for the individual loadcases, you can run another optimization where you run all the loadcases as part of the optimisation but set those compliances from the initial runs as constraints on those loadcases while minimizing volfrac as the objective (the resulting volfrac will be above 30%, but you then know each case is as good as if it were optimized for with 30% material)

Hi Paul, how is your suggestion of running the load cases separately different to running them all in the same study? (With a compliance index response on multiple load steps).

PaulAltair

Combined Compliance Index either didn't exist or was very new when I wrote that reply (it was only introduced with version 2022.0, or at least it wasn't documented before that), honestly I'd not heard of it until you just mentioned it! (and I haven't tried it... yet).

From the description in the manual, if I understand it correctly, it seems that it attempts to automate the process I described without the pain of running the steps individually, by assigning a normalisation factor to the subcase, one difference would still be that it is 'weighting' the NORM (normalisation) of the compliances based on the first iteration (when you have more material).

My 'old' method of getting compliances for all the subcases independently optimised and using those optimised compliances to manually calculate the weightings for a weighted compliance run allows you weight them based on the final 'optimised' compliances for each case.

It would be interesting to compare the results of the 2 approaches!

Paul

D_21643

Combined Compliance Index either didn't exist or was very new when I wrote that reply (it was only introduced with version 2022.0, or at least it wasn't documented before that), honestly I'd not heard of it until you just mentioned it! (and I haven't tried it... yet).

From the description in the manual, if I understand it correctly, it seems that it attempts to automate the process I described without the pain of running the steps individually, by assigning a normalisation factor to the subcase, one difference would still be that it is 'weighting' the NORM (normalisation) of the compliances based on the first iteration (when you have more material).

My 'old' method of getting compliances for all the subcases independently optimised and using those optimised compliances to manually calculate the weightings for a weighted compliance run allows you weight them based on the final 'optimised' compliances for each case.

It would be interesting to compare the results of the 2 approaches!

Paul

Thanks for your reply, I'll let you know how it goes!