modal parameter estimation

[rmunderw]

Hello Everyone,

When using the poly-reference least squares complex frequency domain estimator (p-LSCF) to estimate modal parameters, you iteratively increase the model order and plot successive pole estimates on a stability diagram. In the literature it says you can only increase the model order in steps of Ni (number of inputs). Does anyone know why that is true?

For example:

RMFD - Right Matrix Fraction Description.

Is there a better place to post this question?

Regards,

 

[GregLocock]

You are in the right forum. I'd guess it has to do with how the matrix of results is laid out, and optimised. As disadvantages go it is a pretty trivial one. I've only used polyref a few times, decades ago, so I'm not at all familiar with its ins and outs. Real modal data from a full vehicle, which is what I tended to do, is so scrappy that we didn't waste much time on the niceties of analysis.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[rmunderw]

Thanks Greg,

You can construct the stability diagram by summing all outputs over all references. Or, you can sum over all outputs grouped by all references.

For example the former (sum over all outputs over all references):

The upper limit of the sum is NoNi (# of outputs, # of references)

The T, S, R matrices are combinations of matrices X and Y (X is a shape function, Y is FRF data)

For example the latter (sum over all outputs):

The upper limit of the sum is No (# of outputs)

However, the Y matrix is different in this case (kronecker product)

The Ho matrix has dimensions 1 x Ni (# of references)


Back to the original post and this long winded reply, I believe, as you have said, the dimensions of the matrices in the latter example, are a function of the number of references, the formal example they are not.