Hi All
I am using Reponse surface optimization module ( and not Direct optimization) to optimize my model ( mono component , shell , large deformation, static load, static structural analysis, around 6 parameters)
what are the weaknesses I should expect from this method, just a DOE analysis took half a day for linear analysis, should I forget about nonlinear static analysis ?
if I optimize a component for its maximum stress or displacement capacity, will it be optimized for all other combinations of movements ?
DoE based RSM strategy is a powerful tool to optimize and reduce the amount of experiment (or numerical evaluation) and then time saving, etc. However, if your model is not fitted for a 2nd order model (RSM), the obtained optimal value will not correspond the true one. Before you perform a RSM-FEA, I strongly recommend you to perform a sensitivity analysis using ANOVA or DoE by factorial analysis in order to obtain the main design parameters and see any curvature.
Hello Farzad Torabi ,
regardless of the optimization method I would suggest doing a sensitivity analysis at first to potentially reduce your amount of parameters.
This will also help you ascertain the robustness of your evaluation criteria which you will need to perform an optimization based on the response surface. That's easier said than done, but you could imagine the process as follows:
By the DOE you have already performed the response surface is created to "make sense" of how your criteria change in respect to your altered parameters. The surface should mostly go through your design points. A quality criterium is the COP (Coefficient of Prognosis), you want this to be as close to 100% as possible. Any sudden jumps in your criteria based on the parameters will worsen the quality as your DOE is very unlikely to accurately resolve this region.
Optimizations are often done on nonlinear systems, so I don't see how that would be a problem.
To answer your last question: I don't really understand what you mean by stress and displacement capacity. You should set up seperate simulations for your load cases and define your evaluation criteria appropriately.
Regards
Gino Mathlouthi
Ansys optimization tool is what shall be and the mentioned weakness when revealed isessentially due to the optimization algorithm itself.
Indeed, the response surface is e.g. a "good" alternative when the objective and constraint functions/analysis are very fast or/and the computation of their sensitivities is not available or require too high effort in programming.
Optimizing a component for something (e.g. maximum stress or displacement capacity), does not necessarly optimize for any other objective, and in general it is to expect that optimizing to something may worse everythingelse that was not constrained.
If interested in a simple study on using response surface in comparison to other alternatives (with source codes and APDL macros), I recommend one older proceeding (2006) Conference Paper TECHNIQUES ADOPTED IN INTEGRATING STRUCTURAL ANALYSIS WITH N...
DoE based RSM strategy is a powerful tool to optimize and reduce the amount of experiment (or numerical evaluation) and then time saving, etc. However, if your model is not fitted for a 2nd order model (RSM), the obtained optimal value will not correspond the true one. Before you perform a RSM-FEA, I strongly recommend you to perform a sensitivity analysis using ANOVA or DoE by factorial analysis in order to obtain the main design parameters and see any curvature.
Hello, I advise you to use the superimposed contour graph proposed by minitabe (in response surfaces), you can see in my articles I have already published articles where I used this method.
Cordially;
Hicham
- Badges
- Science topic