Modal analysis of structures under non-stationary excitation

• This work presents a modal analysis based on the adaptive filter RLS/QRD.
• The analysis is performed in signals from structures with rotating machines.
• The filter allows identifying the modal parameters in non-stationary conditions.
• A model of ASCE benchmark structure was used to simulate a numerical case.

The search for mathematical–numerical models to be applied in the design or optimization of civil structures may require adjustment and validation based on experimental data. However, in many cases, when performing dynamic assessment in field structures, it is common to face the situation where it is impossible to stop the running of certain equipment. In this case, all measurements performed on the building and the neighboring structures will be corrupted by noise generated during the operation of these machines. Clearly, many parameter estimation algorithms used to characterize the dynamic behavior of a structure, which are based on the assumption of prior knowledge of the input force in their formulations, may give poor results. In this context, a modal analysis procedure based on the adaptive filter RLS/QRD is described. In this analysis, the measured vibration signal is corrupted by a harmonic noise, in particular, in the presence of signals from the actual machine or other nearby machines. The specific characteristics of the modeling and the hypotheses associated with the harmonic signals are explained. The fact that the excitation force is considered not to be time-invariant in the model, but the frequency varies over time, allows the modal parameters to be identified even under non-stationary conditions. In order to investigate the influence of harmonic noise on the estimated values, a numerical model of the ASCE benchmark structure was used to simulate some practical situations.