Abstract:The proper orthogonal decomposition (POD) -- a popular projection-based model order reduction (MOR) method -- may require significant model dimensionalities to successfully capture a nonlinear solution manifold resulting from a parameterised quasi-static solid-mechanical problem. The local basis method by Amsallem et al. [1] addresses this deficiency by introducing a locally, rather than globally, linear approximation of the solution manifold. However, this generally successful approach comes with some limitations, especially in the data-poor setting. In this proof-of-concept investigation, we instead propose a graph-based manifold learning approach to nonlinear projection-based MOR which uses a global, continuously nonlinear approximation of the solution manifold. Approximations of local tangents to the solution manifold, which are necessary for a Galerkin scheme, are computed in the online phase. As an example application for the resulting nonlinear MOR algorithms, we consider simple representative volume element computations. On this example, the manifold learning approach Pareto-dominates the POD and local basis method in terms of the error and runtime achieved using a range of model dimensionalities.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the limitations of existing Model Order Reduction (MOR) methods in dealing with nonlinear solution manifolds in quasi - static solid mechanics problems. Specifically: 1. **Limitations of the POD method**: - Proper Orthogonal Decomposition (POD), a commonly used projection - based MOR method, may require a relatively large model dimension to successfully capture nonlinear solution manifolds. - For parameterized quasi - static solid mechanics problems, the POD method may not be able to accurately represent complex nonlinear solution manifolds at a lower dimension. 2. **Limitations of the local basis method**: - The local basis method proposed by Amsallem et al. improves the global linear approximation by introducing local linear approximations, but this method performs poorly in the case of scarce data. - The local basis method requires sufficient data to construct an accurate POD ROM in each cluster, and in applications and regions with sparse data, this may lead to inaccurate local linear approximations. - Handling the transitions between clusters is also a challenge and may lead to an unstable or inaccurate scheme. To solve these problems, this paper proposes a graph - based manifold learning method for nonlinear projection MOR. This method uses a global continuous nonlinear approximation to approximate the solution manifold and calculates the local tangents of the solution manifold in the online phase. Compared with the POD and local basis methods, this method has advantages in terms of error and running time, especially when using different model dimensions. ### Specific objectives - Propose a new graph - based manifold learning method for nonlinear projection MOR, especially in quasi - static solid mechanics problems. - Verify the effectiveness of this method through simple Representative Volume Element (RVE) calculations as an example application. - Provide a more efficient and accurate method for approximating the solution manifold in the case of scarce data. ### Key innovation points - Use graph - based manifold learning techniques to obtain an approximation of the global nonlinear solution manifold. - Calculate the local tangents of the solution manifold in the online phase to meet the requirements of the Galerkin scheme. - The proposed method is superior to the traditional POD and local basis methods in terms of error and running time. Through these improvements, the paper aims to provide a more general and efficient reduced - order model method for multiscale problems in solid mechanics.

A manifold learning approach to nonlinear model order reduction of quasi-static problems in solid mechanics

Fast Model Order Reduction Via Nonlinear Optimization

Learning physics-based reduced-order models from data using nonlinear manifolds

Model Order Reduction for High Dimensional Linear Systems Based on Rank-1 Incremental Proper Orthogonal Decomposition.

A local basis approximation approach for nonlinear parametric model order reduction

Model order reduction methods for geometrically nonlinear structures: a review of nonlinear techniques

Model reduction on manifolds: A differential geometric framework

Symplectic Model Reduction of Hamiltonian Systems on Nonlinear Manifolds

Quadratic Approximation Manifold for Mitigating the Kolmogorov Barrier in Nonlinear Projection-Based Model Order Reduction

Scalable nonlinear manifold reduced order model for dynamical systems

Model Order Reduction by Proper Orthogonal Decomposition

Data-free Non-intrusive Model Reduction for Nonlinear Finite Element Models via Spectral Submanifolds

Component based model order reduction with mortar tied contact for nonlinear quasi-static mechanical problems

Finite element based model order reduction for parametrized one-way coupled steady state linear thermomechanical problems

Randomized low‐rank approximation methods for projection‐based model order reduction of large nonlinear dynamical problems

Model Reduction via Dynamic Mode Decomposition

Model reduction of dynamical systems on nonlinear manifolds using deep convolutional autoencoders

Nonlinear Galerkin Method for Dimension Reduction Using Proper Orthogonal Decomposition

Data-driven model reduction via non-intrusive optimization of projection operators and reduced-order dynamics

Nonlinear model order reduction for problems with microstructure using mesh informed neural networks