User Tools

Site Tools


docs:classes:flowfield

This is an old revision of the document!


A PCRE internal error occured. This might be caused by a faulty plugin

====== FlowField ====== The FlowField class represents vector-valued fields of the form <latex> {\bf u}({\bf x}) &= \sum_{k_x,k_y,k_z,j} \hat{u}_{k_x k_y k_z j}\bar{T}_{k_y}(y) \; e^{2 \pi i (k_x x/L_x + k_z z/L_z)} {\bf e}_j </latex> and also scalar and tensor fields with appropriate changes in the dimensionality of the coefficients. The barred %%T%% function is a Chebyshev polynomial scaled to fit the domain y ∈ [a,b]. ((<latex> \bar{T}_{k_y}(y) = T_{k_y}\left(\frac{2}{b-a}(y - \frac{b+a}{2})\right)</latex>)) The spatial domain of a FlowField is Ω = [0,Lx] x [a,b] x [0,Lz], with periodicity in x and z. In channelflow programming, fields such as velocity, pressure, stress tensors, vorticity, etc. are stored as variables of type FlowField. The main functionality of the FlowField class is * algebraic and differential operations, +/-, +=, ∇, ∇<sup>2</sup>, norms, inner products, etc. * transforming back and forth between spectral coefficients <latex> \hat{u}_{k_x k_y k_z j}</latex> and gridpoint values <latex>u_j (x_{n_x}, y_{n_y}, z_{n_z})</latex> * serving as input to DNS algorithms, which map velocity fields forward in time: u(x,t) → u(x, t+Δt) * setting and accessing scpetral coefficients and gridpoint values * reading and writing to disk ===== Constructors / Initialization ===== FlowFields are initialized with gridsize and cellsize parameters, read from disk, or assigned from computations. Examples: <code c++> FlowField f; // null value, 0-d field on 0x0x0 grid FlowField u(Nx, Ny, Nz, Nd, Lx, Lz, a, b); // Nd-dim field on Nx x Ny x Nz grid, [0,Lx]x[a,b]x[0,Lz] FlowField g(Nx, Ny, Nz, Nd, 2, Lx, Lz, a, b); // Nd-dim 2-tensor FlowField h("h"); // read from file "h.ff" FlowField omega = curl(u); </code> ===== Algebraic and differential operators ===== Assume f,g,h etc. are FlowField variables with compatible cell and grid sizes. Examples of possible operations <code> f += g; // f = f + g f = curl(g); f = lapl(g); f = div(g); f = diff(g, j, n); // f_i = d^n g_i /dx_j f = grad(g); // f_ij = dg_i / dx_j f = cross(g,h); f *= 2.7; // f = 2.7*f Real c = L2IP(f,g); // L2 inner product of f,g Real n = L2Norm(u); Real D = dissipation(u); Real E = energy(u); Real I = wallshear(u); </code> The latter functions are defined as <latex> $ \begin{align*} L2IP(f,g) &= \frac{1}{L_x L_y L_z} \int_{\Omega} {\bf f} \cdot {\bf g} \,\, d{\bf x} \\ L2Norm(u) &= \left(\frac{1}{L_x L_y L_z} \int_{\Omega} \|{\bf u}\|^2\,\, d{\bf x} \right)^{1/2}\\ E(u) &= \frac{1}{2 L_x L_y L_z} \int_{\Omega} \|{\bf u}\|^2\,\, d{\bf x} \\ D(u) &= \frac{1}{L_x L_y L_z} \int_{\Omega} \|\nabla \times {\bf u}\|^2 \,\, d{\bf x} \\ I(u) &= \frac{1}{L_x L_z} \int_{y=a,b} \frac{\partial u}{\partial y} \, dx dz \end{align*} $ </latex> ===== Transforms and data access ===== FlowField transforms are a complicated subject --there are transforms in x,y, and z and implicit symmetries in complex spectral coefficients due to the real-valuedness of the field, for instance. This section outlines the bare essentials of transforms and data access methods. For further details see the {{docs:chflowguide.pdf|Channelflow User Guide}}.

docs/classes/flowfield.1234813480.txt.gz · Last modified: 2009/02/16 11:44 by gibson