cs_gradient_boundary.c File Reference

Gradient reconstruction at boundaries and associated functions. More...

#include "cs_defs.h"
#include <assert.h>
#include <errno.h>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <math.h>
#include <float.h>
#include "bft_error.h"
#include "bft_mem.h"
#include "bft_printf.h"
#include "cs_ext_neighborhood.h"
#include "cs_field.h"
#include "cs_halo.h"
#include "cs_internal_coupling.h"
#include "cs_gradient_priv.h"
#include "cs_math.h"
#include "cs_mesh.h"
#include "cs_mesh_adjacencies.h"
#include "cs_mesh_quantities.h"
#include "cs_parall.h"
#include "cs_gradient_boundary.h"

Include dependency graph for cs_gradient_boundary.c:

Functions
void	cs_gradient_boundary_iprime_lsq_s (const cs_mesh_t *m, const cs_mesh_quantities_t *fvq, const cs_internal_coupling_t *cpl, cs_lnum_t n_faces, const cs_lnum_t *face_ids, cs_halo_type_t halo_type, double clip_coeff, const cs_real_t *bc_coeff_a, const cs_real_t *bc_coeff_b, const cs_real_t c_weight[], const cs_real_t var[], cs_real_t *restrict var_iprime)
	Compute the values of a scalar at boundary face I' positions using least-squares interpolation. More...
void	cs_gradient_boundary_iprime_lsq_v (const cs_mesh_t *m, const cs_mesh_quantities_t *fvq, const cs_internal_coupling_t *cpl, cs_lnum_t n_faces, const cs_lnum_t *face_ids, cs_halo_type_t halo_type, double clip_coeff, const cs_real_t *bc_coeff_a[3], const cs_real_t *bc_coeff_b[3][3], const cs_real_t c_weight[], const cs_real_t var[][6], cs_real_t *restrict var_iprime[3])
	Compute the values of a vector at boundary face I' positions using least-squares interpolation. More...
void	cs_gradient_boundary_iprime_lsq_t (const cs_mesh_t *m, const cs_mesh_quantities_t *fvq, const cs_internal_coupling_t *cpl, cs_lnum_t n_faces, const cs_lnum_t *face_ids, cs_halo_type_t halo_type, double clip_coeff, const cs_real_t *bc_coeff_a[6], const cs_real_t *bc_coeff_b[6][6], const cs_real_t c_weight[], const cs_real_t var[][6], cs_real_t *restrict var_iprime[6])
	Compute the values of a symmetric tensor at boundary face I' positions using least-squares interpolation. More...

Detailed Description

Gradient reconstruction at boundaries and associated functions.

Function Documentation

cs_gradient_boundary_iprime_lsq_s()

void cs_gradient_boundary_iprime_lsq_s	(	const cs_mesh_t *	m,
		const cs_mesh_quantities_t *	fvq,
		const cs_internal_coupling_t *	cpl,
		cs_lnum_t	n_faces,
		const cs_lnum_t *	face_ids,
		cs_halo_type_t	halo_type,
		double	clip_coeff,
		const cs_real_t *	bc_coeff_a,
		const cs_real_t *	bc_coeff_b,
		const cs_real_t	c_weight[],
		const cs_real_t	var[],
		cs_real_t *restrict	var_iprime
	)

Compute the values of a scalar at boundary face I' positions using least-squares interpolation.

This assumes ghost cell values for the variable (var) are up-to-date.

A simple limiter is applied to ensure the maximum principle is preserved (using non-reconstructed values in case of non-homogeneous Neumann conditions).

Remarks

To compute the values at I', we only need the gradient along II', so in most cases, we could simply assume a Neumann BC for a given face.

We still use the provided BC's when possible, for the following cases:

• Given a non-uniform Dirichlet condition and a non-orthogonal mesh, the Dirichlet values at face centers (shifted by II' relative to I') can convey a portion of the information of the gradient along II'.
• For cells with multiple boundary faces, information from faces whose normals are not orthogonal to II' can also provide a significant contribution to the normal.

For coupled faces (with internal coupling), we assume a Neumann BC, as the values that could be computed at the intersections of the segments joining coupled cell centers and the matching boundary face do not currently account for the non-linearity associated with wall laws.

A more precise computation for cells with multiple coupled boundary faces would be possible by using multiple passes, first estimating boundary face values without recontruction, then using that information for faces which are not tangential to II' in a second pass. Note though that as the mesh is refined, the portion of the boundary adjacent to cells with multiple coupled faces (i.e. at corners and ridges) diminishes.

Parameters

[in]	m	pointer to associated mesh structure
[in]	fvq	pointer to associated finite volume quantities
[in]	cpl	structure associated with internal coupling, or NULL
[in]	n_faces	number of faces at which to compute values
[in]	face_ids	ids of boundary faces at which to compute values, or NULL for all
[in]	halo_type	halo (cell neighborhood) type
[in]	clip_coeff	clipping (limiter) coefficient (no limiter if < 0)
[in]	bc_coeff_a	boundary condition term a, or NULL
[in]	bc_coeff_b	boundary condition term b, or NULL
[in]	c_weight	cell variable weight, or NULL
[in]	var	variable values et cell centers
[out]	var_iprime	variable values et face iprime locations

cs_gradient_boundary_iprime_lsq_t()

void cs_gradient_boundary_iprime_lsq_t	(	const cs_mesh_t *	m,
		const cs_mesh_quantities_t *	fvq,
		const cs_internal_coupling_t *	cpl,
		cs_lnum_t	n_faces,
		const cs_lnum_t *	face_ids,
		cs_halo_type_t	halo_type,
		double	clip_coeff,
		const cs_real_t *	bc_coeff_a[6],
		const cs_real_t *	bc_coeff_b[6][6],
		const cs_real_t	c_weight[],
		const cs_real_t	var[][6],
		cs_real_t *restrict	var_iprime[6]
	)

Compute the values of a symmetric tensor at boundary face I' positions using least-squares interpolation.

This assumes ghost cell values which might be used are already synchronized.

A simple limiter is applied to ensure the maximum principle is preserved (using non-reconstructed values in case of non-homogeneous Neumann conditions).

This function uses a local iterative approach to compute the cell gradient, as handling of the boundary condition terms b in higher dimensions would otherwise require solving higher-dimensional systems, often at a higher cost.

Remarks

To compute the values at I', we only need the gradient along II', so in most cases, we could simply assume a Neuman BC.

The same logic is applied as for cs_gradient_boundary_iprime_lsq_s.

Parameters

[in]	m	pointer to associated mesh structure
[in]	fvq	pointer to associated finite volume quantities
[in]	cpl	structure associated with internal coupling, or NULL
[in]	n_faces	number of faces at which to compute values
[in]	face_ids	ids of boundary faces at which to compute values, or NULL for all
[in]	halo_type	halo (cell neighborhood) type
[in]	clip_coeff	clipping (limiter) coefficient (no limiter if < 0)
[in]	bc_coeff_a	boundary condition term a, or NULL
[in]	bc_coeff_b	boundary condition term b, or NULL
[in]	c_weight	cell variable weight, or NULL
[in]	var	variable values et cell centers
[out]	var_iprime	variable values et face iprime locations

cs_gradient_boundary_iprime_lsq_v()

void cs_gradient_boundary_iprime_lsq_v	(	const cs_mesh_t *	m,
		const cs_mesh_quantities_t *	fvq,
		const cs_internal_coupling_t *	cpl,
		cs_lnum_t	n_faces,
		const cs_lnum_t *	face_ids,
		cs_halo_type_t	halo_type,
		double	clip_coeff,
		const cs_real_t *	bc_coeff_a[3],
		const cs_real_t *	bc_coeff_b[3][3],
		const cs_real_t	c_weight[],
		const cs_real_t	var[][6],
		cs_real_t *restrict	var_iprime[3]
	)

Compute the values of a vector at boundary face I' positions using least-squares interpolation.

This assumes ghost cell values which might be used are already synchronized.

A simple limiter is applied to ensure the maximum principle is preserved (using non-reconstructed values in case of non-homogeneous Neumann conditions).

This function uses a local iterative approach to compute the cell gradient, as handling of the boundary condition terms b in higher dimensions would otherwise require solving higher-dimensional systems, often at a higher cost.

Remarks

To compute the values at I', we only need the gradient along II', so in most cases, we could simply assume a Neuman BC.

The same logic is applied as for cs_gradient_boundary_iprime_lsq_s.

Parameters

[in]	m	pointer to associated mesh structure
[in]	fvq	pointer to associated finite volume quantities
[in]	cpl	structure associated with internal coupling, or NULL
[in]	n_faces	number of faces at which to compute values
[in]	face_ids	ids of boundary faces at which to compute values, or NULL for all
[in]	halo_type	halo (cell neighborhood) type
[in]	clip_coeff	clipping (limiter) coefficient (no limiter if < 0)
[in]	bc_coeff_a	boundary condition term a, or NULL
[in]	bc_coeff_b	boundary condition term b, or NULL
[in]	c_weight	cell variable weight, or NULL
[in]	var	variable values et cell centers
[out]	var_iprime	variable values et face iprime locations