The role of anti-resonance frequencies from operational modal analysis in finite element model updating

Finite element model updating traditionally makes use of both resonance and modeshape information. The mode shape information can also be obtained from anti-resonance frequencies, as has been suggested by a number of researchers in recent years. Anti-resonance frequencies have the advantage over mode shapes that they can be much more accurately identified from measured frequency response functions. Moreover, anti-resonance frequencies can, in principle, be estimated from output-only measurements on operating machinery. The motivation behind this paper is to explore whether the availability of anti-resonances from such output-only techniques would add genuinely new information to the model updating process, which is not already available from using only resonance frequencies. This investigation employs two-degree-of-freedom models of a rigid beam supported on two springs. It includes an assessment of the contribution made to the overall anti-resonance sensitivity by the mode shape components, and also considers model updating through Monte Carlo simulations, experimental verification of the simulation results, and application to a practical mechanical system, in this case a petrol generator set. Analytical expressions are derived for the sensitivity of anti-resonance frequencies to updating parameters such as the ratio of spring stiffnesses, the position of the centre of gravity, and the beam's radius of gyration. These anti-resonance sensitivities are written in terms of natural frequency and mode shape sensitivities so their relative contributions can be assessed. It is found that the contribution made by the mode shape sensitivity varies considerably depending on the value of the parameters, contributing no new information for significant combinations of parameter values. The Monte Carlo simulations compare the performance of the update achieved when using information from: the resonances only; the resonances and either anti-resonance; and the resonances and both anti-resonances. It is found that the addition of anti-resonance information improves the updating performance for some combinations of parameter values, but does not improve the update in significant other regions. The simulated results are verified using resonance and anti-resonance frequencies measured on a steel beam test rig. The investigation is extended to include the updating of parameters of a petrol generator set. It is found that the contribution of the anti-resonances to the model update is heavily dependent on the geometry of the model and the choice of variables to be updated, suggesting that, for some models, the pursuit of anti-resonance information through expensive operational modal analysis may be inappropriate.