Frequency resolution and number of averages effects on the accuracy and dispersion of operational modal analysis results

Structural Health Monitoring techniques based on the study of the evolution of modal parameters aim to detect a structural modification relying on the results of a modal parameter estimation algorithm. Basically, a change in modal parameter values may indicate a structure evolution and possibly an ongoing damage situation. Beside structure changes there are other reasons that can cause changes in modal parameter values. Environmental and operational conditions naturally cause a variation of modal parameters. In addition the results of identification are affected by a dispersion due to the signal processing method. This paper is entirely focused on this latter aspect. The study exploits a Monte Carlo method relying upon the modal model of the G. Meazza stadium grandstands in Milan, where a permanent monitoring system is working since several years. Thanks to the modal model the response of the structure to random excitation is simulated. Numerical data are analyzed with the same algorithm used to process real data coming from the sensors that measure the dynamic response of the structure. Since the algorithm takes as input the power spectra associated to the acceleration responses of the monitored structure, modal parameter estimations are heavily affected by sub-record length, that is to say frequency resolution and number of averages used to compute power spectra. The work specifically discussed in the paper aims to investigate the separate effects of these two parameters on accuracy and spread of the estimated eigenfrequencies and non-dimensional damping ratios. Identification process spread is a key factor in assessing the real potential of the monitoring technique since represents a threshold under which is not possible to recognize a modal parameter change linked to a structural modification. Outcomes are discussed taking into account the complexity of the modal analysis performed on the proposed test case structure whose dynamic behaviour is characterized by the presence of close and highly damped modes.