Dear Burak,

MAC matrices are also defined for complex eigenvectors.

Best regards,    

                                  Nils

MAC COMAC is a good way

MAC COMAC can handle complex eigenvalues as well.

Thanks for all answers! I already did the experiment (EMA) with 120 points. The shape is like cylinder. On the other hand, i modeled in Abaqus. With modal updating, i will update material parameters (E modul, density....) according to frequence and MAC value. Problem is now in sensivity analyses some paremeter are already zero effects and there is no any paremeter to fulfill all the freqency and MAC.

I try to investigate the problem now; a) because of the real-complex mode shaps b) measurment error c) wrong FEM model? Do you have any experience about this?

By the way, in CoMAC, I see some points has bad effects on the FRF function (low MAC Matrix). Should i eliminate these points and make FRF fucntion again?

Dear Burak,

The idealization of boundary conditions in your finite element model is prone to errors.

Are the any joints in your test structure ?

Best regards,

                               Nils

Hi Nils,

My test structure is a stator and there is no joints in the structure. In FEM Model, I simulated without any B.C (or free B.C) with Modal Analyses Modul (abaqus).

How did you support your test structure (e.g. by soft springs etc.) ?

Yes, I support my test structure with soft springs.

HI Burak ,

If Mode shape extraction is your problem , try frequency domain correlation like FRAC or FDAC , which will correlate the FRF directly.

orbi.ulg.be/bitstream/2268/25147/1/via4.pdf

with FRAC you will compare complex FRF matrix values directly.

Hi Burak,

Please let me know, if you can share your Abaqus model (*.inp).

Hi Pratik!

Actually my problem is; I m doing Modal updating and I do not know which mode shapes (real or complex) should I use it. Because in FEM Model (without damping), real mode is calculated. But in Modal Analyse (EMA), mode shapes are complex (because of damping). Which one should use or it effects the result?

On the other hand, In Abaqus I do not use B.C, should I add B.C?

Thanks again

Hi Nils,

unfortunately no. But it is cylinder shape, there are also some other parts which are connected to cylinder (with tie constraint).

Hi , Burak

For validation of your model you should use Real mode of FEM with Real mode EMA

refer : http://sem-proceedings.com/13i/sem.org-IMAC-XIII-13th-13-13-4-Extracting-Real-Modes-from-Complex-Measured-Modes.pdf

Alternatively you can also correlate converting both in Phase-Mag form and compare Magnitude

Also as first step My method was to corelate the undampned eigen frequencies spatially and then adjust damping later ( you can use suitable optimisation procedure to do this.)

Please note that the size of your tied regions have an effect on the modeshapes. Can you guarantee that there are no relative motions between the different parts of the assembly ?

My structure is Stator (together with windings). I defined tie constraint (abaqus) in between stator and windings. There are also end windings.

Can you share some pictures of your experimental set-up and the finite element model ?

Hi Burak, in Friswell and Mottershead book about FE model updating, says that the complex modes shall be transformed to equivalent real modes to compare with the analitycal real modes.

Hi Burak

You get & should have complex modes when damping is localized in your system, e.g. when acoustic-structure domains couple, single elements with one dashpot damper etc.

Complex modes require both damping modelling and complex solvers. Both are more difficult than real modes and assuming light and evenly distribuyed damping, i.e. real modes.

You can get 'false' complex modes in your test also from things like

• not having measured on 'the whole system'
• non linearities, typically from gap - these can be detected using recirpocity, just swap excitation/response positions and measure FRFs again and compare Acc1/Force2 with Acc2/Force1.
• Massloading when your accelerometer is moved and heavy enough to affect the dynamic response of your measurement object. .

Take a look here for more ramblings on the subject

http://qringtech.com/learnmore/why-simulate-measure-correlate-automate/

http://qringtech.com/2014/06/22/designed-damping-types-mechanisms-application-limitation/

Sincerely

Claes