Gradient Operator Dotted with Normal Vector

[TLWise]

This is a different spin to a previous question to satisfy my need to code to the PDE formulation for the problem directly in MOOSE. Given the equation (I decided to derive it in PDE form) represented in the picture, is there a already defined post processor that will provide this solution?

[GiudGiud]

Hello

I dont know about one. I dont think this would be very difficult to implement. You should try and report on your progress and difficulties here

[GiudGiud]

Can you paste the header as well ?

The type of the property must not be consistent with using the multiplication with *

[TLWise]

Sure thank you. Here it is.:

'''
#pragma once

// MOOSE includes
#include "SideIntegralVariablePostprocessor.h"

// Forward Declarations
template <bool, typename>
class SideDiffusiveFluxIntegral_crabTempl;
typedef SideDiffusiveFluxIntegral_crabTempl<false, Real> SideDiffusiveFluxIntegral_crab;
typedef SideDiffusiveFluxIntegral_crabTempl<true, Real> ADSideDiffusiveFluxIntegral_crab;
typedef SideDiffusiveFluxIntegral_crabTempl<false, RealVectorValue> SideVectorDiffusivityFluxIntegral;
typedef SideDiffusiveFluxIntegral_crabTempl<true, RealVectorValue> ADSideVectorDiffusivityFluxIntegral;

/**

• This postprocessor computes a side integral of the mass flux.
  */
  template <bool is_ad, typename T>
  class SideDiffusiveFluxIntegral_crabTempl : public SideIntegralVariablePostprocessor
  {
  public:
  static InputParameters validParams();

  SideDiffusiveFluxIntegral_crabTempl(const InputParameters & parameters);

protected:
Real computeQpIntegral() override;

const GenericMaterialProperty<T, is_ad> * const _diffusion_coef;
/// Whether the normal component has been selected
const bool _use_normal;

/// Will hold 0, 1, or 2 corresponding to x, y, or z.
const int _component;
/// Whether an advected variable was supplied in the input
const bool _advected_variable_supplied;
/// Whether a diffusion coefficient variable was supplied in the input
const bool _diffusion_coef_supplied;
/// Variable storing the advected quantity; used for finite elements
const VariableValue & _advected_variable;

/// Velocity components
const Moose::Functor<Real> & _vel_x;
const Moose::Functor<Real> * const _vel_y;
const Moose::Functor<Real> * const _vel_z;

//private:
// /// Routine to get the diffusive flux with a Real diffusivity
// RealVectorValue diffusivityGradientProduct(const RealVectorValue & grad_u, Real diffusivity);
//
// /// Routine to get the diffusive flux with a RealVectorValue diffusivity
// RealVectorValue diffusivityGradientProduct(const RealVectorValue & grad_u,
// const RealVectorValue & diffusivity);
}; ''

[GiudGiud]

_diffusion_coef is a pointer so you need to do

(*_diffusion_coef)[_qp])

to dereference it

[TLWise]

Here is the new error:

''' error: cannot convert 'libMesh::boostcopy::enable_if_c<true, libMesh::TypeVector >::type' {aka 'libMesh::TypeVector'} to 'libMesh::Real' {aka 'double'} in return
93 | return (_use_normal ? raw_value((_diffusion_coef)[_qp]) *
| ~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| |
| libMesh::boostcopy::enable_if_c<true, libMesh::TypeVector >::type {aka libMesh::TypeVector}
94 | RealVectorValue(vel_x, vel_y, vel_z) * _normals[_qp]
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
95 | : raw_value((_diffusion_coef)[_qp]) *
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
96 | RealVectorValue(vel_x, vel_y, vel_z)(_component)); '''

[GiudGiud]

right now you have 3 vectors multiplying each other, which does not end up with a single value like you want

D is a vector
normal is vector
RealVectorValue(vel_x, vel_y, vel_z) is a vector

I think you need to take the component of one these or something

[TLWise]

Got it. I will remove one to work towards a solution. Is it possible to feed the result into a vector that stores the result? I still need to apply grad_u to rhe velocity vector?

[GiudGiud]

What stands out most to me is that the max pressure in the FV model is almost twice that of the FEM solution.

Is the pressure drop correct though? As in is this max pressure just a local maxima?
Is the mesh converged? And are both solutions not showing any oscillation?

[TLWise]

Both solutions demonstrate no oscillation. The pressure drop is established by an analytical inlet velocity function and a zero pressure outlet boundary conditions.

[GiudGiud]

And you have twice the pressure drop in FV.
the definition of the friction factors are different in both discretizations
Are you using our friction kernels ?

or is the pressure drop only from the no slip on the channel walls?

[GiudGiud]

can you please share both input files? maybe we can spot what's wrong

[TLWise]

The pressure drop is only from the no slip boundary condition. I will send you the files needed for each.

[TLWise]

Thank you for taking a look at this for me. Let me know if I left something out that prevents you from running the code. T. Leon Wise

…

________________________________ From: Guillaume Giudicelli ***@***.***> Sent: Sunday, November 24, 2024 7:48 AM To: idaholab/moose ***@***.***> Cc: Wise, Tammy ***@***.***>; Author ***@***.***> Subject: Re: [idaholab/moose] Gradient Operator Dotted with Normal Vector (Discussion #28566) can you please share both input files? maybe we can spot what's wrong — Reply to this email directly, view it on GitHub<#28566 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AJQ6AYQGCD2SH5624PYZFJT2CHYNXAVCNFSM6AAAAABNZOWGBKVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCMZWGM3TEMQ>. You are receiving this because you authored the thread.