This paper provides an analysis on the effects of exact and inexact integrations on stability, convergence, numerical diffusion, and numerical oscillations for the Eulerian-Lagrangian method (ELM). In the finite element ELM, when more accurate integrations are used for the right-hand-side, less numerical diffusion is introduced and better approximation is obtained. When linear interpolation is used for numerical integrations, the resulting ELM is shown to be unconditionally stable and of first-order accuracy. When Gauss quadrature is used, conditional stability and second-order accuracy are established under some mild constraints for the convection-diffusion problems. Finally, numerical experiments demonstrate that more accurate integrations lead to better approximation, and spatial adaptivity can substantially reduce numerical oscillations and smearing that often occur in the ELM when inexact numerical integrations are used.

This paper presents a fast algorithm (BS2 Algorithm) for fitting $C^1$ surfaces to scattered data points. By using energy minimization, the bivariate spline space $S^1_2(∆^{(2)}_{m,n})$ is introduced to construct a $C^1$-continuous piecewise quadratic surface through a set of irregularly 3D points. Moreover, a multilevel method is also presented. Some experimental results show that the accuracy is satisfactory. Furthermore, the BS2 Algorithm is more suitable for fitting surfaces if the given data points have some measurement errors.

We consider a new subgrid eddy viscosity method based on pressure projection and extrapolated trapezoidal rule for the transient Navier-Stokes equations by using lowest equal-order pair of finite elements. The scheme stabilizes convection dominated problems and ameliorates the restrictive inf-sup compatibility stability. It has some attractive features including parameter free for the pressure stabilized term and calculations required for higher order derivatives. Moreover, it requires only the solutions of the linear system arising from an Oseen problem per time step and has second order temporal accuracy. The method achieves optimal accuracy with respect to solution regularity.

In this paper, a complex parameter is employed in the Hermitian and skew-Hermitian splitting (HSS) method (Bai, Golub and Ng: SIAM J. Matrix Anal. Appl., 24(2003), 603-626) for solving the complex linear system $Ax=f$. The convergence of the resulting method is proved when the spectrum of the matrix $A$ lie in the right upper (or lower) part of the complex plane. We also derive an upper bound of the spectral radius of the HSS iteration matrix, and an estimated optimal parameter $α$(denoted by $α_{est}$) of this upper bound is presented. Numerical experiments on two modified model problems show that the HSS method with $α_{est}$ has a smaller spectral radius than that with the real parameter which minimizes the corresponding upper bound. In particular, for the 'dominant' imaginary part of the matrix $A$, this improvement is considerable. We also test the GMRES method preconditioned by the HSS preconditioning matrix with our parameter $α_{est}$.

A novel collocation method for a coupled system of singularly perturbed linear equations is presented. This method is based on rational spectral collocation method in barycentric form with sinh transform. By sinh transform, the original Chebyshev points are mapped into the transformed ones clustered near the singular points of the solution. The results from asymptotic analysis about the singularity solution are employed to determine the parameters in this sinh transform. Numerical experiments are carried out to demonstrate the high accuracy and efficiency of our method.

In this work, a new numerical scheme is proposed for thermal/isothermal incompressible viscous flows based on operator splitting. Unique solvability and stability analysis are presented. Some numerical results are given, which show that the proposed scheme is highly efficient for the thermal/isothermal incompressible viscous flows.