UPDATING ANALYTICAL MODELS BY USING LOCAL AND GLOBAL PARAMETERS AND RELAXED OPTIMISATION REQUIREMENTS

Abstract Two methods aimed at improving the robustness of the identification process in the presence of more or less unavoidable idealisation and measurement errors are described. The first method splits the uncertain parameters into two groups. The first group contains the local physical parameters related to those areas of the structure (called the main structure), where parameter uncertainties can be assigned a priori; the second group contains global generalised parameters related to the remaining structure (the residual structure), where it is difficult to assign uncertain parameters. These global parameters compensate for all the effects resulting from non-parametric modeling errors. This approach has the additional advantage of restricting the measurement of the mode shapes to those critical areas where local parameter uncertainties are expected. The second method describes a technique for smoothing the experimental mode shapes in order to reduce the influence of random and systematic measurement errors. Constructing the objective function from the differences between the analytical and the smoothed experimental mode shapes together with the related eigenfrequencies relaxes the minimisation requirements and tends to improve the robustness of the identification process. The applicability of these methods is demonstrated by updating the parameters related to bolted joints of an experimental frame structure.