Parameter optimisation
issueid=367 01-10-2011 04:21 AM
Junior Member
Parameter optimisation

I have started using the FEBio optimisation to characterise porous, hyperelastic scaffold material, using the Ogden material and have had good results. Although the current implementation is sufficient for uni-axial tests (the only data I have available at this point in time) it would be great if the optimisation procedure could be expanded.

Aspects that I thought of would be to have various models for various test setup, and then to have a global optimisation function. Also to include multiple features in a single model, for instance optimising for both the tensile force and cross section strain during a uni-axial test.

This expansion would help significantly with characterising anisotropy, and will also help with determining bulk modulus for highly comressible materials.

EDIT: Adding a scenario for bi-axial testing would also be great, thus two force and displacement plots for one model.

Thanks!
Hugo
Issue Details
Issue Number 367
Issue Type Feature
Project FEBio
Category Unknown
Status Implemented
Priority 2
Suggested Version Unknown
Implemented Version (none)
Votes for this feature 0
Votes against this feature 0
Assigned Users (none)
Tags (none)




01-13-2011 11:20 AM
Lead Code Developer
Hi Hugo,

I don't see a problem with your suggestion so we will definitely look into that. I'm not sure when we'll get to it though, since right now we are working on implementing a better optimization algorithm than what is currently available. The current implementation has some severe limitations (e.g. no constraints on parameters) that need to be addressed first.

Cheers,

Steve.

01-14-2011 01:17 AM
Junior Member
Thanks Steve

Yes indeed, I forgot to mention that implementation of limits would be fantastic. What is the expected time to release of the improvements currently being worked on?

Cheers
Hugo

01-14-2011 03:17 PM
Developer
Hi Hugo,

I'm working on implementing the constrained parameter optimization method. I can't give you a firm commitment, but we are talking on the order of weeks.

Cheers,

Dave

01-20-2011 02:26 AM
Junior Member
Hi Dave

That is fantastic.

Thanks
Hugo

05-25-2011 05:03 PM
Developer
Hugo,

The constrained optimization routine in the Intel MKL proved not to be suitable for the our purposes. We have, however, implemented the constrained optimization routine from the NAG library. This requires that you have a NAG license for your computer. We will continue to look for a routine that does not require an additional license.

Cheers,

Dave

05-26-2011 03:36 AM
Junior Member
Hi Dave

Thanks for letting me know. If you can explain what the new optimisation system entails I will be able to decide how much effort to put into trying to find a NAG license here at UCT

Regards
Hugo

05-27-2011 12:39 PM
Developer
Hugo,

You use the same input file that is described in chapter 7 of the FEBio User's manual. We need to do a little testing to make sure that we can compile the code with the NAG library and still have all the other other features work if a user does not have a NAG license. I'll let you know when we have that confirmed. Note that this will only work on Linux 64 bit machines.

Cheers,

Dave

05-27-2011 04:35 PM
Developer
Hugo,

I realized that the input parameters for the NAG library are slightly different from that outlined in the user's manual, but I am updating it to use the same nomenclature. If you want to try it out before I get the changes committed, let me know and I'll send you a sample input file.

Cheers,

Dave

07-13-2015 11:53 AM
Lead Code Developer
Just going through some old posts. I assume this has been resolved. Let me know otherwise.

Cheers,

Steve

+ Reply