M.S. Thesis Presentation by Christopher W. Moloney
Thursday, May 8, 2003

(Dr. Jerry Ginsberg, advisor)

"Visual and Analytical Characteristics for the Identification of Complex Modes"


This study is concerned with identifying modal contributions within very noisy frequency response functions (FRF). Four coherence metrics are proposed to aid in the location of resonant frequencies. The metrics are mean, mean-square, variance, and wavelet decomposition. Each seeks to operate on the FRF data set such that the output is maximized at the frequency of the most dominant mode. The metrics are applied to windowed portions of the FRF data. The windows step through the FRF data covering different frequency intervals. The analysis takes several passes through the FRF data, using progressively shorter window lengths. The entire set of results is analyzed for coherence. The frequency band containing the largest metric is determined for each pass through the data, or for each analysis window length. If the frequency bands of the maximum metric values overlap, then it is assumed modal data is present. If the frequency bands are scattered across the entire frequency range of the FRF data, then the FRF data is deemed to be lacking modal data. The effectiveness and consistency of the coherence metrics are gauged through a series of numerical simulations on fabricated SDOF and MDOF FRFs. Suggestions for further investigation into the coherence metrics as well as the possibility of integrating the metrics into SDOF mode fitting algorithms are then presented.