Thursday, May 14, 2009

FRF matrix - a matrix containing data points from all the FRF measurements you've made on a structure. It would be a 3D matrix where each "slice" of the matrix is a different frequency bin (f1, f2, f3...). Each entry of that "slice" is the value of each FRF at each measurement location. Ideally, this should be a symmetric matrix since m(1,2) is ALWAYS equal to m(2,1), where m(1,2) is the response at location 2 due to excitation at location 1, and m(2,1) is the response at location 1 due to excitation at location 2.

Well conditioned matrix - a matrix whose inverse is insensitive to minor changes in values. Sometimes, when you have FRFs where some of the values are very large (100k G's e.g.) and others are very small (.1G, e.g.), the inverse is very sensitive to changes in those small numbers. Since in testing, those small measurements could be near the noise floor, their values could change arbitrarily, causing problems in the inverse calculation.

Deconvolution - is an algorithm-based process used to reverse the effects of convolution on recorded data. convolution is simply when you have an input that is multiplied by a transfer function to result in a different output. Deconvolution is simply when you have the output and transfer function and work backwards to get the input.