Parameter Adaptation With Automatically Selected Co-Ordinates From Measured Eigenvectors As a Way to a Validated FE Model

Abstract

Abstract This paper considers an algorithm on the basis of parameter adaptation for mass and stiffness embedded in the eigenvalue problem solver. The algorithm is intended for large finite element (FE) models. The errors, which can be reduced by the procedure described in this paper, occur due to detailed features, which would require an unduly fine mesh to be included in the model, or in uncertainties in the description of mechanical behaviour, material properties, etc. Another source for errors are model reduction techniques (superelement technique) necessary for the application of the model structure in an automatic control circuit (smart structures). It is a well-known fact that the natural frequencies can be measured much more accurately than mode shapes, for mode shapes can only be measured for accessible regions and normally for translational degrees of freedom (DOF). Therefore the algorithm uses only measured natural frequencies (frequency differences) and the calculated mode shape vectors to determine the parameter changes. In a new approach it is also possible to select automatically, or by experience, those co-ordinates from the measured mode shape vectors that correspond to points with high sensitivity or other very reliable points. An interface system designed to exchange data between the experimental modal analysis system (EMA) and the FE program ensures, that the measured and calculated mode shape vectors are orthonormalised in the same way and the points of the FE mesh correspond to the pick up points for the measurement. Examples of industrial parts at the end of the paper illustrate how the procedure works and what influence we can obtain by inclusion of some co-ordinates of measured mode shape vectors.