Application of matched field processing methods to source localization in structural dynamics.

摘要

Matched Field Processing (MFP) is a source localization technique that incorporates detailed environmental models into array signal processing techniques. This dissertation investigates the application of MFP to localizing broadband random vibration sources on finite, non-rigid structures. Applications are investigated through theoretical investigation, numerical simulations on a canonical structural dynamics system, and by experiments on a planar welded steel frame. Comparisons are made between conventional Bartlett, adaptive minimum variance and eigenvector beamforming methods for MFP. In addition, pseudo-inverse techniques for input force identification are compared to MFP methods. A review of MFP in underwater acoustics is presented as background, and comparisons to results in the underwater acoustics research are made. Results indicate that MFP is a useful and powerful tool for localizing vibration sources as a means of diagnosing vibration problems in flexible structures.