Efficient compression of reduced impedance matrices for trimmed structural models: a novel methodology and industrial validation

In response to increasingly strict vibration and noise regulations, software frameworks such as Nastran and Actran have stepped up to the challenge, providing sophisticated solutions to efficiently model damping treatments within numerical methods. One of the available techniques is the representation of trim components as frequency-dependent reduced impedance matrices (RIM) in a frequency response analysis of fully trimmed models. While RIM enables coupling between physical or modal-based components, challenges arise due to large dense matrices. Each matrix is reduced either on physical coupling DOFs or modes of vibration of the component and stored on disk. Since each matrix is dense and large in industrial models, it leads to storage and efficiency issues related to computational and input/output (I/O) operations. This paper introduces a methodology to compress RIMs, particularly for modal components, by selecting the most contributing modes to represent the coupling relationship of each surface. Numerical results on an industrial car body model demonstrate that this approach effectively eliminates low-effect contributions, minimizes storage requirements, and reduces computational time. General Motors' involvement in validating the method adds real-world applicability and reliability to the findings, enhancing its practical engineering value.