Numerical integration of homogeneous functions on convex and nonconvex polygons and polyhedra

Eric Chin, Jean B. Lasserre, N. SukumarComputational 2015 · DOI

TL;DR

This paper presents a novel method for accurately and efficiently integrating homogeneous functions over both convex and nonconvex polygons and polyhedra. By leveraging Euler's homogeneous function theorem and Stokes's theorem, the authors reduce complex volume integrals to simpler surface or even point evaluations at vertices, offering a significant computational advantage.

What Problem Does This Paper Address?

Accurate numerical integration is crucial in many computational fields, including finite element methods and virtual element methods, especially when dealing with complex geometries like polygons and polyhedra. Traditional methods often struggle with nonconvex shapes or require computationally expensive tessellations. This paper tackles the challenge of developing a unified and efficient integration technique that works reliably for both convex and nonconvex geometries, and for a broad class of functions known as homogeneous functions.

Methods

The core of the paper lies in the application of two fundamental mathematical theorems:

  • Euler's Homogeneous Function Theorem: This theorem relates a homogeneous function to its gradient and the function's value at a point. The authors use this to simplify integrals.
  • Generalized Stokes's Theorem (Gauss's Divergence Theorem): This theorem allows for the conversion of volume integrals into surface integrals. The authors recursively apply this to reduce the integration dimension.

By combining these, the integration of a homogeneous function over a polytope is reduced to an integral over its boundary facets. For homogeneous polynomials, this process can be further simplified to evaluating the function and its derivatives at the vertices of the polytope. The paper also extends these methods to handle integration over domains bounded by polar curves and to address weakly singular and discontinuous integrands. An efficient algorithm is provided for implementation.

Datasets

Not discussed in the paper. The paper focuses on developing a general numerical method and demonstrates its application with various geometric shapes (polygons and polyhedra) and polynomial functions.

Key Insights and Ideas

  • Unified Approach for Convex and Nonconvex Geometries: A significant contribution is the extension of integration methods to nonconvex polygons and polyhedra, overcoming a limitation of previous work.
  • Reduction to Vertex Evaluation: For homogeneous polynomials, the integration can be reduced to function and derivative evaluations solely at the polytope's vertices, drastically simplifying computations.
  • Handling of Weakly Singular Integrands: The method is shown to be effective for functions that become unbounded at certain points, which is a common challenge in numerical integration.
  • Polar Coordinate Integration: The paper provides a framework for integrating functions in polar coordinates, particularly useful for domains with curved boundaries.

Key Results

The paper demonstrates that the proposed method achieves results that exactly match those obtained from exact integration tools (like LattE) for a wide range of test cases, including:

  • Convex and nonconvex polygons.
  • Simple and nonsimple nonconvex polygons.
  • Convex and nonconvex polyhedra.
  • Weakly singular and discontinuous integrands.

The numerical examples show that the polar-transformed integral often converges more quickly than the Cartesian integral, especially for weakly singular functions.

Limitations and Future Work

  • Implementation Complexity: While the method is theoretically sound, the implementation of calculating hyperplane coefficients ($a_i$ and $b_i$) and rotation matrices can be non-trivial, especially for higher dimensions.
  • Future Work: The authors plan to explore the application of their integration scheme to extended and embedded finite element methods, as well as Galerkin methods on polygons and polyhedra. They also suggest further investigation into the geometric method of degree-reduction using rotations.

Summary updated

BibTeX
@inproceedings{chin2015numerical,
  author = {Eric Chin and Jean B. Lasserre and N. Sukumar},
  title = {{Numerical integration of homogeneous functions on convex and nonconvex polygons and polyhedra}},
  booktitle = {Computational},
  year = {2015},
  doi = {10.1007/s00466-015-1213-7},
}
Figure