Computational study demonstrates structured second-order modeling from frequency domain data, indicating improved physical interpretability over classical methods.
Key Points
To develop data-driven modeling algorithms that preserve second-order differential structures directly from frequency domain measurements.
Extended the Adaptive Antoulas-Anderson (AAA) algorithm to second-order dynamical systems using a structured barycentric form.
Formulated algorithmic variants optimizing either computation speed or approximation accuracy based on computational constraints.
Derived theoretical error and performance bounds and evaluated methods across three numerical benchmarks against unstructured approaches.
Successfully identified second-order dynamical system models directly from frequency response data while maintaining physical structure.
Demonstrated superior modeling effectiveness and physical interpretability compared to classical unstructured data-driven techniques across three benchmark problems.