R. Cavoretto,A. De Rossi
Abstract:In this paper we present an adaptive discretization technique for solving elliptic partial differential equations via a collocation radial basis function partition of unity method. In particular, we propose a new adaptive scheme based on the construction of an error indicator and a refinement algorithm, which used together turn out to be ad-hoc strategies within this framework. The performance of the adaptive meshless refinement scheme is assessed by numerical tests.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to construct an adaptive discretization technique for elliptic partial differential equations (such as Poisson's equation). Specifically, the author proposes an adaptive discretization scheme based on the Radial Basis Function - Partition of Unity Method (RBF - PUM) and evaluates the performance of this scheme through numerical experiments.
### Problem Background
When solving elliptic partial differential equations, mesh - free methods have attracted much attention due to their flexibility and simple implementation for high - dimensional problems. In particular, the Radial Basis Function (RBF) method performs well in geometrically complex problems and has a high order of convergence because of its mesh - free characteristics. However, traditional RBF methods may encounter serious ill - conditioned problems when dealing with large - scale problems, leading to computational instability.
### Research Objectives
The main objective of this paper is to design an adaptive discretization technique for the RBF - PUM method to improve the accuracy and stability of the numerical solution. Specifically, the author proposes an adaptive scheme based on error indicators and refinement algorithms, which can effectively evaluate and improve the quality of the numerical solution without the need for an exact solution.
### Method Overview
1. **RBF - PUM Method**:
- Divide the original domain into several sub - domains and construct a local RBF approximation on each sub - domain.
- Use the global RBF - PUM approximation form \(\tilde{u}(x)=\sum_{j = 1}^{d}w_j(x)\tilde{u}_j(x)\), where \(w_j\) is the weight function on the sub - domain and \(\tilde{u}_j\) is the local RBF approximation.
2. **Adaptive Discretization Technique**:
- **Error Indicators**: Two error indicators are defined. One estimates the error by comparing the difference between the global RBF - PUM solution and the local RBF interpolation solution; the other estimates the error by comparing the difference between two RBF - PUM solutions on different node sets.
- **Refinement Algorithm**: Based on the residual sampling method, adjust the set of discretization points by adding or removing nodes to optimize the quality of the numerical solution.
### Numerical Experiments
To verify the effectiveness of the proposed adaptive scheme, the author carried out a large number of numerical experiments. The experimental results show that the adaptive method can significantly improve the accuracy of the numerical solution and reduce the number of required nodes while maintaining computational stability.
### Main Conclusions
By introducing the adaptive discretization technique, this paper successfully improves the performance of the RBF - PUM method in solving elliptic partial differential equations. Compared with traditional methods, this adaptive scheme not only improves the accuracy of the numerical solution but also reduces the degree of ill - conditioning in the calculation, thereby enhancing the stability and reliability of the method.
### Formula Display
For example, the discrete form of Poisson's equation can be expressed as:
\[
-\Delta\tilde{u}(x_i)=-\sum_{j = 1}^{d}\Delta(w_j(x_i)\tilde{u}_j(x_i)) = f(x_i),\quad x_i\in\Omega
\]
where,
\[
\Delta(w_j(x_i)\tilde{u}_j(x_i))=\Delta w_j(x_i)\tilde{u}_j(x_i)+2\nabla w_j(x_i)\cdot\nabla\tilde{u}_j(x_i)+w_j(x_i)\Delta\tilde{u}_j(x_i)
\]
These formulas show how to improve the numerical solution of the RBF - PUM method through the adaptive discretization technique.
Hopefully, this summary can help you understand the main content and research purpose of this paper. If you have more questions or need further explanation, please feel free to let us know!