On the application of output-only modal identification to base excited frame structures

The work focuses on the application of modal identification techniques to plan-asymmetric building structures that are subjected to base excitation. The challenging problem related to the identification of high-order modes and, especially, torsional modes is taken into account for different lengths of time and directions of the applied base excitation. The evaluation of the identification accuracy was performed through a simulation approach: analytical models of plan-asymmetric RC frames, which are characterized by different structural eccentricities, were subjected to white noise base excitation. This kind of input is rather idealized for real buildings, which are generally excited by earthquake or traffic loads, but it is commonly adopted in experimental programs related to vibration-based damage detection using scaled or full-size buildings tested on shaking tables. The output-only modal identification was applied using the response time-histories only and the modal parameters, in terms of natural circular frequencies, modal damping ratios and mode shapes, were determined. These identified parameters were then compared with the true values related to the initially assumed analytical models. Firstly, the analyses were performed for one structural configuration, for a fixed direction of the applied base excitation, but considering various durations of such input. The results showed that the accuracy related to mode shapes identification was lower for the first torsional mode with respect to the second longitudinal mode, which is a higher-order mode. In addition, the identified torsional mode shape exhibited major modal complexity than the longitudinal mode, as a result of simulated noise and identification errors. Secondly, the analyses were conducted considering different structural configurations, a fixed duration of the base excitation, but different directions of the applied input. The outcomes showed that the accuracy in mode shape estimations, evaluated by calculating the Modal Assurance Criterion (MAC) between identified and analytical mode shapes, was strongly dependent on the direction of the applied input, especially for high-order modes. It was also observed that these MAC values are directly correlated with the modal participation factors of the structure, evaluated for the different directions of the base excitation. Referring to the torsional mode shapes, the results showed that the number of input directions, for which these modes can be identified, are higher for the structures with major structural eccentricities.