Keywords: large-scale analysis, error measures, vibroacoustics, wind turbines.

The development and analysis of large-scale finite element models is an important step in engineering product development in order to obtain reliable and accurate solutions of the related physical problem. It is a particular challenge when either structures are very large and consist of many single components that are connected at interfaces, or when dynamic analysis is involved where a sufficient fine mesh resolution is required depending on the underlying frequency range. In this paper we develop implementable and practically useful error measures that are actually applicable for general large-scale problems. For evaluating the required mesh density depending on the actually considered frequency space of the problem, methods from experimental dynamics are employed within a pure computational framework such as the model assurance criterion (MAC-value) and the frequency response assurance criterion (FRAC-value). By employing these criteria it will be shown that the error in frequencies and mode shapes can be easily reduced to a desired level of accuracy. In order to deal with large-scale problems, standard reduction order methods like the component mode synthesis (CMS) method are employed to drastically reduce the computational effort in the analysis of the full model. To this end, a new practical approach is proposed where the components are not separated at the actual interfaces but at virtual interfaces within the component itself in order to overcome the difficulties of modelling of complex interfaces with (possibly nonlinear) solution behavior. As a result of our approach we found that the definition of virtual interfaces within certain components leads to overall results that are up to 50 % more accurate compared to the classical approach where the CMS interface is located at the actual interfaces of the component. The methodologies presented in this paper are developed for the vibroacoustic analysis of a 2.5 MW wind turbine at large scale. Despite the special focus on wind turbines, the error estimation procedure presented here is not restricted to applications from structural dynamics. Once a good quality of the discrete models has been established, they can be used to obtain reliable and accurate results also in other largescale engineering applications. In the considered examples, we find that the procedure is very practical, easy to use and can be easily computed with data from any finite element code.

download the full-text of this paper (PDF, 6 pages, 416 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description