南开大学计算物理讲义9.pptVIP

b）隐格式 利用关系 可得三层隐式格式： 截断误差 : 绝对稳定 9.3.4 Model for general initial conditions Our analysis of the Euler method for solving the diffusion equation in section 9.3.3 assumed initial conditions of the form sin(kπx/Lx) sin(lπy/Ly) where k, l are integers and Lx, Ly are the dimensions of the domain. In addition to satisfying the boundary conditions, these initial conditions represent a set of orthogonal functions which may be used to construct any arbitrary initial conditions as a Fourier series. Now, since the diffusion equation is linear, and as our stability analysis of the previous section shows the conditions under which the solution for each Fourier mode is stable, we can see that the equation (178) applies equally for arbitrary initial conditions. 9.3.5 Implicit methods The implicit method is significantly better in terms of stability than the Euler method for ordinary differential equations. For partial differential equations such as the diffusion equation we may analyze this in the same manner as the Euler method of section 9.3.3. 9.3.5.1 Backward Euler 9.3.5.2 Crank-Nicholson Backward Euler Yn = Yn-1+Δt f(Yn,tn) Yn+1 = Yn+Δt f(Yn +1,tn +1) Crank-Nicholson 9.3.5.1 Backward Euler Yn = Yn-1+Δt f(Yn,tn) Yn+1 = Yn+Δt f(Yn +1,tn +1) r=2+s (x,t-k) (x+h,t) (x,t) (x-h,t) r=4 9.3.5.2 Crank-Nicholson μ = Δt/Δx2 for the i=1,m-1 internal points. Solution of this expression will involve the solution of a tridiagonal system for this one-dimensional problem at each time step: (x,t) (x+h,t-k) (x,t-k) (x-h,t-k) (x-h,t) (x+h,t) To test the stability we again choose a Fourier mode. Here we have only one spatial dimension so we use U(0) i = sin(θi) which satisfies the boundary condition if θ is a multiple of π. Substituting this into (182) we find [1–2μsin2(θ/2)]/[1+2μsin2(θ/2)] 1 for all μ 0 with u(x=0,t) = u(x=1,t) = 0 and 9.4 Hyperbolic Equations - Wave Equations advection equation 一阶线性双曲型方程最简单的形式为 当给定初始条件 9.4.1. The 1-d advection equation we discretize in time on th