# 4. Advection Stabilization¶

In general, advection for both the edge and element-based scheme is identical with standard exception of the location of the integration points. The full advection term is simply written as,

where \(\phi\) is \(u_i\), \(Z\), \(h\), etc.

The evaluation of \(\phi_{ip}\) defines the advection stabilization choice. In general, the advection choice is a cell Peclet blending between higher order upwind (\(\phi_{upw}\)) and a generalized un-stabilized central (Galerkin) operator, \(\phi_{gcds}\),

In the above equation, \(\eta\) is a cell Peclet blending. The generalized central operator can take on a pure second order or pseudo fourth order form (see below). For the classic Peclet number functional form (see Equation (4.3)) a hybrid upwind factor, \(\gamma\), can be used to ensure that no stabilization is added (\(\eta = 0\)) or that full upwind stabilization is included (as will be shown, even with limiter functions). The hybrid upwind factor allows one to modify the functional blending function; values of unity result in the normal blending function response in Figure Fig. 4.1; values of zero yield a pure central operator, i.e., blending function of zero; values \(>>\) unity result in a blending function value of unity, i.e., pure upwind. The constant \(A\) is implemented with a value of 5. The value of this constant can not be changed via the input file. Note that this functional form is named the “classic” form within the input file.

The classic cell Peclet blending function is given by

The classic Peclet functional form makes it difficult to dial in the exact point at which the Peclet factor transitions from pure upwind to pure central. Therefore, an alternative form is provided that has a hyperbolic tangeant functional form. This form allows one to specify the transition point and the width of the transition (see Equation (4.4)). The general tanh form is as follows,

Above, the constant \(c_{trans}\) represents the transition Peclet number while \(c_{width}\) represents the width of the transition. The value of \(\lambda\) is simply the shift between of the raw tanh function from zero while \(\delta\) is the difference between the max Peclet factor (unity) and the minimum value prior to normalization. This approach ensures that the function starts at zero and asymptotes to unity,

The cell-Peclet number is computed for each sub-face in the element from the two adjacent left (L) and right (R) nodes,

A dot-product is implied by repeated indices.

The upwind operator, \(\phi_{upw}\) is computed based on a blending of the extrapolated state (using the projected nodal gradient) and the linear interpolated state. Second or third order upwind is provided based on the value of \(\alpha_{upw}\) blending

The extrapolated value based on the upwinded left (\(\phi^L\)) or right (\(\phi^R\)) state,

The distance vectors are defined based on the distances between the L/R points and the integration point (for both edge or element-based),

In the case of all transported quantities, a Van Leer limiter of the extrapolated value is supported and can be activated within the input file (using the solution options “limiter” specification).

Second order central is simply written as a linear combination of the nodal values,

where \(N^{ip}_k\) is either evaluated at the subcontrol surface or edge midpoint. In the case of the edge-based scheme, the edge midpoint evaluation provides for a skew symmetric form of the operator.

The generalized central difference operator is provided by blending with the extrapolated values and second order Galerkin,

where,

The value of \(\alpha\) provides the type of psuedo fourth order stencil and is specified in the user input file.

The above set of advection operators can be used to define an idealized one dimensional stencil denoted by (\(i-2\), \(i-1\), \(i\), \(i+1\), \(i+2\)), where \(i\) represents the particular row for the given transported variable. Below, in the table, the stencil can be noted for each value of \(\alpha\) and \(\alpha_{upw}\).

\(i-2\) |
\(i-1\) |
\(i\) |
\(i+1\) |
\(i+2\) |
\(\alpha\) |
\(\alpha_{upw}\) |

\(0\) |
\(-\frac{1}{2}\) |
\(0\) |
\(+\frac{1}{2}\) |
\(0\) |
\(0\) |
n/a |

\(+\frac{1}{8}\) |
\(-\frac{6}{8}\) |
\(0\) |
\(+\frac{6}{8}\) |
\(-\frac{1}{8}\) |
\(\frac{1}{2}\) |
n/a |

\(+\frac{1}{12}\) |
\(-\frac{8}{12}\) |
\(0\) |
\(+\frac{8}{12}\) |
\(-\frac{1}{12}\) |
\(\frac{2}{3}\) |
n/a |

\(+\frac{1}{4}\) |
\(-\frac{5}{4}\) |
\(+\frac{3}{4}\) |
\(+\frac{1}{4}\) |
\(0\) |
\(\dot m > 0\) |
\(1\) |

\(0\) |
\(-\frac{1}{4}\) |
\(-\frac{3}{4}\) |
\(+\frac{5}{4}\) |
\(-\frac{1}{4}\) |
\(\dot m < 0\) |
\(1\) |

\(+\frac{1}{6}\) |
\(-\frac{6}{6}\) |
\(+\frac{3}{6}\) |
\(+\frac{2}{6}\) |
\(0\) |
\(\dot m > 0\) |
\(\frac{1}{2}\) |

\(0\) |
\(-\frac{2}{6}\) |
\(-\frac{3}{6}\) |
\(+\frac{6}{6}\) |
\(-\frac{1}{6}\) |
\(\dot m < 0\) |
\(\frac{1}{2}\) |

It is noted that by varying these numerical parameters, both high quality, low dissipation operators suitable for LES usage or limited, monotonic operators suitable for RANS modeling can be accomodated.