The Frequency Domain Decomposition (FDD) is a newly developed output only modal analysis technique that identifies the natural frequencies, damping ratios and mode shapes of a dynamic system from measured response data. This thesis describes an investigation into the capabilities of the FDD as a modal identification tool. The primary reason for this investigation is to identify the benefits and limitations of the technique and assess its usefulness in modal analysis.
A study of the FDD identification algorithm was first conducted to gain an understanding of how the algorithm extracts modal parameters from response data. The identification algorithm was then implemented in the programming language MATLAB and applied to response data obtained from the simulation of a 2 DOF mass-spring-damper system to illustrate how the algorithm works.
The identification of a more complex 6 DOF mass-spring-damper system was then conducted to further investigate the capabilities of the FDD identification algorithm. Three cases were considered for the simulation of the 6 DOF system. The first case involved the simulation of the system with light damping and no noise in the response data. The second case considered involved the simulation of the system with the level of system damping varied from light to heavy, to investigate the ability of the technique to estimate damping. The third case involved the simulation of the system with various levels of noise added to the response data to determine the sensitivity of the technique to noise. In each case, the FDD identification algorithm was applied to the simulated data and the results obtained were validated against the exact system modal parameters determined from an eigenvalue analysis.
The results of the identification of the 6 DOF system indicated that the FDD is capable of producing highly accurate natural frequency, damping ratio and mode shape estimates subject to the assumptions that the input excitation is white noise and the system is lightly damped. It was also found that the technique is insensitive to noise in the response data. Even with high levels of noise contamination of the response signals, the FDD technique identified the natural frequencies and mode shapes with a high degree of accuracy. The quality of the damping estimates produced by the technique was found to decline for higher levels of system damping and noise in the response data. The FDD technique was also found to be capable of identifying closely spaced modes. In the 6 DOF system analysis, the technique was able to identify the two pairs of close modes in the system irrespective of the level of system damping or noise in the response data.
The FDD was found to be a powerful output only modal analysis technique. Besides producing accurate modal parameter estimates of lightly damped dynamic systems, the FDD is capable of identifying closely spaced modes, is robust to noise and is user friendly.