A meshless method based on the moving least squares (MLS) approximation for the numerical solution of two-dimensional nonlinear integral equations of the second kind on non-rectangular domains (Q466865)

This paper investigates a numerical method for solving two-dimensional nonlinear Fredholm integral equations of the second kind on non-rectangular domains: \[ u(x,t)-\lambda\int_{D} K(x,t,y,s,u(y,s))dsdy=f(x,t), \quad (x,t)\in D. \] where the right hand side function \(f\) and the kernel function \(K\) are given, \(u(x,t)\) is the unknown function to be determined, \(\lambda\) is a constant and \(D\subset \mathbb{R}^2\) is a non-rectangular bounded region. The scheme utilizes the shape functions of the moving least squares (MLS) approximation constructed on scattered points as a basis in the discrete collocation method. The MLS methodology is an effective technique for approximating unknown functions which involves a locally weighted least squares polynomial fitting. The proposed method is meshless, since it does not need any background mesh or cell structures and so it is independent of the geometry of the domain. The scheme reduces the solution of two-dimensional nonlinear integral equations to the solution of systems of nonlinear algebraic equations. In order to approximate double integrals on the domain \(D\) the composite Gauss-Legendre-type quadrature formulas are used in the scheme. The error analysis of the proposed method is provided. The efficiency and accuracy of the new technique are illustrated by several numerical examples.