Systematic determination of bias and scatter in different experimental modal analysis methods

Kurzfassung

Experimental modal analysis is a tool to identify mathematical substitute models from measured vibrations for applications such as response prediction, controller design, and model validation. Since vibration measurements are affected by noise, the quality of the identified substitute models depends on the signal-to-noise ratio. Noise can cause stochastic errors (scatter) and systematic errors (bias) in the estimated modal parameters. This paper investigates the extent of scatter and bias with noisy input data for three representative modal analysis methods: the Least-Squares Complex Frequency-Domain (LSCF) estimator, the Least-Squares Complex Exponential (LSCE) estimator, and the Eigensystem Realization Algorithm (ERA). These methods represent different frequency- and time-domain identification approaches. Bias and scatter of the identified modal parameters (i.e. eigenfrequency, damping ratio, residue, modal mass) are evaluated systematically with noise for these methods in a Monte Carlo study using an analytical system.