TL;DR
This paper introduces a novel "immersed isogeometric analysis with boundary conformal quadrature" (IBC-IGA) method for simulating complex 2D elasticity problems, especially those involving large deformations. By precisely integrating over trimmed elements using re-parameterized B-spline patches, the method achieves high accuracy and efficiency without the need for complex meshing, outperforming existing immersed methods.
What Problem Does This Paper Address?
Simulating complex engineering structures with finite element methods (FEM) often requires extensive and time-consuming pre-processing, particularly for generating meshes that conform to intricate geometries. Immersed and fictitious domain methods offer a way around this by embedding the physical domain within a simpler background mesh, but they face challenges in accurately integrating over "trimmed" elements (those cut by the physical boundary). This integration accuracy is crucial, especially for nonlinear problems like finite deformation elasticity, where numerical instability can arise. The authors aim to improve the accuracy, robustness, and efficiency of these immersed methods by developing a more sophisticated integration technique for trimmed elements.
Methods
The core innovation is the boundary conformal quadrature applied within an immersed isogeometric analysis (IGA) framework.
- Isogeometric Analysis (IGA): Instead of traditional FEM meshes, IGA uses the same spline basis functions (like B-splines or NURBS) used in CAD to represent both the geometry and the solution field. This allows for exact representation of complex curves and surfaces.
- Immersed Approach: The physical domain (Ω₀) is embedded within a simpler, larger background domain (Ω), typically a Cartesian grid of IGA elements. The physical domain is defined by its boundary representation (B-rep).
- Boundary Conformal Quadrature: For elements cut by the physical boundary (trimmed elements), the authors employ a three-step process:
- Approximation: The boundary segment within a trimmed element is approximated by a Bézier curve of the same polynomial degree as the spline basis.
- Decomposition: The trimmed element is decomposed into active (inside Ω₀) and inactive (outside Ω₀) parts. Complex shapes are further subdivided.
- Re-parameterization: The active parts of the trimmed elements are re-parameterized using B-spline surfaces. Classical Gauss quadrature is then applied to these re-parameterized patches, ensuring accurate integration over the physical domain.
- Handling Fictitious Domain: To manage the non-physical domain (Ωf), a small penalty factor (α) is applied to integrals outside Ω₀.
- Deformation Map Resetting: For large deformation problems, to prevent numerical instabilities arising from extreme deformations in the fictitious domain, the deformation gradient is reset to identity for integration points in Ωf at each Newton-Raphson iteration.
- Weak Boundary Condition Enforcement: Essential (Dirichlet) boundary conditions are imposed weakly using Nitsche's method (for linear elasticity) or a penalty method (for nonlinear elasticity).
Datasets
Not discussed in the paper. The paper uses benchmark problems with known analytical solutions or reference solutions from other numerical methods for validation.
Key Insights and Ideas
- Precise Integration is Key: The accuracy of immersed methods hinges on how well trimmed elements are integrated. The proposed boundary conformal quadrature, by re-parameterizing trimmed regions, achieves near-exact integration without refining the mesh itself.
- IGA's Advantage in Immersed Methods: The smooth, high-order nature of IGA basis functions, combined with the ability to represent complex geometries exactly, makes it a natural fit for immersed methods, especially when dealing with trimmed elements.
- Deformation Map Resetting is Crucial for Stability: For finite deformation problems, the fictitious domain can become highly deformed, leading to instability. Resetting the deformation map in this domain significantly improves robustness without affecting the accuracy in the physical domain.
- Efficiency Through Reduced Quadrature Points: The boundary conformal quadrature requires significantly fewer integration points compared to traditional hierarchical or adaptive methods, especially at higher polynomial degrees, leading to faster simulations.
Key Results
- Optimal Convergence Rates: The method demonstrates optimal convergence rates (h- and k-refinement) for both linear and nonlinear elasticity problems.
- High Accuracy: The L₂-norm of the displacement error and strain energy error are shown to be very low, often reaching machine precision for well-posed problems.
- Robustness for Large Deformations: The combination of boundary conformal quadrature, fictitious domain penalization, and deformation map resetting allows for stable and accurate simulations of finite deformation elasticity problems.
- Efficiency Gains: The IBC-IGA method significantly reduces the number of quadrature points needed compared to methods like hierarchical quadtree (HQT) integration, leading to orders of magnitude fewer computational operations for integration.
- Successful Application to Homogenization: The method is successfully applied to the homogenization of microstructured materials under finite deformations, demonstrating its utility for multiscale simulations.
- Handling Complex Geometries: The method handles geometries with closely spaced inclusions and multiple kinks effectively, showcasing its robustness.
Limitations and Future Work
- 2D Focus: The current work is limited to 2D problems.
- Penalty Parameter Tuning: While the method is robust, the choice of penalty parameters (for the fictitious domain and boundary conditions) can still influence accuracy and convergence, requiring careful selection or further investigation.
- Future Work: The authors suggest extending the approach to 3D geometries, applying it to more complex multiscale problems (e.g., 3D printing), and exploring multiphysics applications. They also note the need for further investigation into optimal penalty parameter values.
Summary updated