Efficient broad-band simulation of large scale vibroacoustic systems with random input data

• Applicants: Professor Dr. Matthias Bollhöfer; Professorin Dr.-Ing. Sabine C. Langer; Professor Dr.-Ing. Ulrich Römer
• Subject Area: Acoustics
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 531569940

Large-scale vibroacoustic analysis is at the core of modern acoustic design efforts. Despite of the maturity of the finite element method, a broad-band analysis of strongly coupled problems still poses computational challenges. The situation becomes more involved because of the need to quantify uncertainties, which is crucial for resilience to environmental and manufacturing variability. The project aims to address these issues with a hybrid approach, combining sparse surrogate and reduced order modeling in an error-controlled, adaptive way and is founded on a set of well-chosen test cases from building and vehicle acoustics, with high-dimensional parametric uncertainty. Large-scale finite element analysis with modern iterative solvers is used to generate data in the parameter space as largescale numerical linear algebra methods and adaptivity are key drivers of further automatization of computer codes for vibroacoustics. Effective surrogate and reduced order models are selected based on (nonlinear) quantitites of interest and error measures, which inform acoustic design decisions. The core concept of the computational methodology is a hybrid approach, which employs a reduced basis and a surrogate model to handle the frequency and random parameter dependencies, respectively. Error estimators and cost models allow to optimally allocate computational resources within the hybrid approach. Moreover, multi-fidelity data, generated by varying the mesh resolution and solver accuracies of the finite element models, are employed to further enhance the efficiency of the hybrid method. The project addresses structure preserving concepts and the re-use of local sampling information and data acquired during solver calls. Overall this project will make a significant contribution to surrogate and reduced order modeling for vibroacoustics and dynamical systems, well beyond the state-of-the-art, addressing realistic and challenging benchmark examples from the field.

• DFG Programme: Research Grants