Evaluating the Effect of Modelling Errors in Load Identification Using Classical Identification Methods

Abstract

Load identification, or input identification as the more general term, is a field of study that requires a wide set of disciplines, which suffers from uncertainties caused by the challenges within each discipline. When making load identification, several different approaches exist. For all (or at least most) methods, however, some sort of system model is required. This model may be simple or complex, depending on the system at hand. Typically, if the identification process is vibration fed, the system model will be created from modal parameters. These parameters, however, are often subject to uncertainty and thus may be considered as stochastic variables. In this paper, the root causes of uncertainty for load identification are demonstrated using classical identification techniques. From a numerical perspective, uncertainty is quantified through Monte Carlo simulations. Two results are outlined: one where the identification process is completely blindfolded in its most naive form, and one where the spatial distribution of the load is predefined. In general, it is found that fixing the spatial distribution of the load can compensate for truncation errors in the modal parameters.