A local basis approximation approach for nonlinear parametric model order reduction

Konstantinos Vlachas, Konstantinos Tatsis, Konstantinos Agathos, Adam R. Brink, Eleni N. ChatziJournal of Sound and Vibration 2020 · DOI · OpenReview

TL;DR

This paper introduces a novel approach to Parametric Model Order Reduction (pMOR) for complex, nonlinear structural systems. By developing a new method for interpolating local basis functions, they achieve accurate predictions of system behavior under varying material properties and excitation conditions, significantly speeding up simulations.

What Problem Does This Paper Address?

Modern engineering systems, from buildings to wind turbines, are increasingly complex and operate under dynamic, often nonlinear, conditions. Accurately simulating their behavior requires high-fidelity models, but these are computationally expensive, especially when considering variations in material properties (like damage or wear) or external forces (like earthquakes). This computational burden hinders real-time monitoring, diagnostics, and control. The challenge is to create reduced-order models (ROMs) that are both accurate across a range of parameters and computationally efficient, capturing the "as-is" state of a system.

Methods

The core of this work lies in a physics-based parametric Model Order Reduction (pMOR) approach. They leverage Proper Orthogonal Decomposition (POD) to extract dominant modes from simulation snapshots of the system's nonlinear dynamic response. The key innovation is a new local basis interpolation technique. Instead of interpolating the entire basis vectors, they interpolate a reduced coefficient matrix that maps local bases to a global one. This is performed on the tangent space of the Grassmann manifold to preserve orthogonality properties. This approach is combined with hyper-reduction techniques (specifically, the Energy-Conserving Sampling and Weighting - ECSW method) to further accelerate the evaluation of nonlinear terms.

Datasets

The paper uses two primary case studies for validation:

  • A two-story building with nonlinear links: This academic example uses a Bouc-Wen model for material nonlinearity, with parametric dependency on the hysteresis amplitude and a related parameter.
  • A 3D wind turbine tower: This more complex, real-world example is subjected to earthquake excitation. The parametric dependency here lies in the amplitude and frequency content of the earthquake spectrum.

No new datasets were introduced; rather, existing models and simulated data were used for evaluation.

Key Insights and Ideas

  • Interpolating coefficients is more efficient than interpolating basis vectors: The proposed method significantly reduces computational cost by interpolating a smaller coefficient matrix, rather than the full basis vectors, especially for high-dimensional systems.
  • Local bases are crucial for nonlinear systems: A single global basis struggles to capture the complex, localized nonlinear behavior that arises from varying parameters. Local bases, tailored to specific parameter regions, offer better accuracy.
  • Manifold interpolation preserves essential properties: Projecting local bases onto the tangent space of the Grassmann manifold is critical for maintaining orthogonality and ensuring stable interpolation.
  • Hyper-reduction is essential for practical speedup: While pMOR reduces the model's dimension, hyper-reduction is necessary to tackle the computational bottleneck of evaluating nonlinear terms, leading to significant speedups.

Key Results

The proposed pMOR approach demonstrated:

  • High accuracy: The method achieved low relative errors (e.g., below 3% for restoring forces in the two-story building example and below 4% for stress contours in the wind turbine tower) across various parametric configurations.
  • Significant speedup: The hyper-reduced pROM achieved speedups of up to 38.8x compared to the High Fidelity Model (HFM) for the wind turbine tower case study, while maintaining good accuracy.
  • Effective handling of material nonlinearity and excitation variability: The method successfully captured the dynamic response and stress states for systems with material hysteresis and for earthquake excitations with varying spectral characteristics.
  • Comparison of strategies: The paper showed that the proposed coefficients interpolation method generally outperformed direct entries interpolation and local basis approaches, especially for finer subdomain partitioning.

Limitations and Future Work

  • Limited parameter range: The current implementation was validated for a maximum of two parameters. For multi-parameter scenarios, more sophisticated sampling and interpolation methods are needed.
  • Subdomain definition: The definition of subdomains and the clustering of snapshots based on similarity are crucial for accuracy and currently rely on manual partitioning or simplified assumptions.
  • Generalization of earthquake parameterization: Future work could explore more generalized parameterizations of earthquake time histories.

The authors suggest that further research is needed to address these limitations and to develop more generalized parameterization techniques for broader applicability.

Summary updated

Abstract

The efficient condition assessment of engineered systems requires the coupling of high fidelity models with data extracted from the state of the system `as-is'. In enabling this task, this paper implements a parametric Model Order Reduction (pMOR) scheme for nonlinear structural dynamics, and the particular case of material nonlinearity. A physics-based parametric representation is developed, incorporating dependencies on system properties and/or excitation characteristics. The pMOR formulation relies on use of a Proper Orthogonal Decomposition applied to a series of snapshots of the nonlinear dynamic response. A new approach to manifold interpolation is proposed, with interpolation taking place on the reduced coefficient matrix mapping local bases to a global one. We demonstrate the performance of this approach firstly on the simple example of a shear-frame structure, and secondly on the more complex 3D numerical case study of an earthquake-excited wind turbine tower. Parametric dependence pertains to structural properties, as well as the temporal and spectral characteristics of the applied excitation. The developed parametric Reduced Order Model (pROM) can be exploited for a number of tasks including monitoring and diagnostics, control of vibrating structures, and residual life estimation of critical components.
BibTeX
@article{vlachas2020local,
  author = {Konstantinos Vlachas and Konstantinos Tatsis and Konstantinos Agathos and Adam R. Brink and Eleni N. Chatzi},
  title = {{A local basis approximation approach for nonlinear parametric model order reduction}},
  journal = {Journal of Sound and Vibration},
  year = {2020},
  doi = {10.1016/j.jsv.2021.116055},
}
Figure