As vibration and noise regulations become more stringent, numerical models need to incorporate more detailed damping treatments. Commercial frameworks, such as Nastran and Actran, allow the representation of trim components as frequency-dependent reduced impedance matrices (RIM) in direct frequency response (DFR) analysis of fully trimmed models.
The RIM is versatile enough to couple the trims to modal-based or physical components. If physical, the trim components are reduced on the physical coupling degrees of freedom (DOFs) for each connected interface. If modal, the RIMs are projected on the eigenmodes of the connected component. While a model size reduction is achieved compared to the original model, most numerical models possess an extensive number of interfaces DOFs, either modal or physical, resulting in large, dense RIMs that demand substantial memory and disk storage. Thus, the approach faces challenges related to storage capacities and efficiency, because of the demanding computational input/output (I/O) operations involved.
This paper introduces a new robust and efficient methodology. It aims to further compress these RIMs when dealing with modal components. Instead of performing a conventional modal projection, the method reduces the global modes onto the coupling surfaces of each component to their most significant contributions. The paper demonstrates, on an industrial fully trimmed car body model, that if the truncation process eliminates low-effect contributions sufficiently, the coupling is adequately represented, resulting in a significant reduction in disk storage with minimal loss of accuracy. As an additional benefit, the computational time is reduced due to the I/O handling of much smaller matrices.