Example problem: 2D driven cavity flow in a quarter-circle domain with spatial adaptation.

In this example we shall demonstrate

  • how easy it is to adapt the code for the solution of the driven cavity problem in a square domain, discussed in a previous example , to a different domain shape,
  • how to apply body forces (e.g. gravity) in a Navier-Stokes problem,
  • how to switch between the stress-divergence and the simplified forms of the incompressible Navier-Stokes equations.


The example problem

In this example we shall illustrate the solution of the steady 2D Navier-Stokes equations in a modified driven cavity problem: The fluid is contained in a quarter-circle domain and is subject to gravity which acts in the vertical direction. We solve the problem in two different formulations, using the stress-divergence and the simplified form of the Navier-Stokes equations, respectively, and by applying the gravitational body force via the gravity vector, $ {\bf G} $, and via the body force function, $ {\bf B}$, respectively.

Problem 1:
The 2D driven cavity problem in a quarter circle domain with gravity,
using the stress-divergence form of the Navier-Stokes equations

Solve

\[ Re\phantom{i}u_j\frac{\partial u_i}{\partial x_j} = - \frac{\partial p}{\partial x_i} + \frac{Re}{Fr} \, G_i + \frac{\partial }{\partial x_j} \left( \frac{\partial u_i}{\partial x_j} + \frac{\partial u_j}{\partial x_i} \right), \ \ \ \ \ \ \ \ \ \ (1) \]

and

\[ \frac{\partial u_i}{\partial x_i} = 0, \]

in the quarter-circle domain $ D = \{x_1 \geq 0 $, $ x_2 \geq 0 $ and $ x_1^2+x_2^2 \leq 1 \} $, subject to the Dirichlet boundary conditions

\[ \left. \mathbf{u}\right|_{\partial D}=(0,0), \ \ \ \ \ \ \ \ \ \ (2) \]

on the curved and left boundaries; and

\[ \left. \mathbf{u}\right|_{\partial D}=(1,0), \ \ \ \ \ \ \ \ \ \ (3) \]

on the bottom boundary, $ x_2 = 0 $. Gravity acts vertically downwards so that $ (G_1, G_2) = (0,-1) $.


When discussing the implementation of the Navier-Stokes equations in an earlier example , we mentioned that oomph-lib allows the incompressible Navier-Stokes equations to be solved in the simplified, rather than the (default) stress-divergence form. We will demonstrate the use of this feature by solving the following problem:

Problem 2:
The 2D driven cavity problem in a quarter circle domain with gravity,
using the simplified form of the Navier-Stokes equations

Solve

\[ Re\phantom{i}u_j\frac{\partial u_i}{\partial x_j} = - \frac{\partial p}{\partial x_i} + B_i + \frac{\partial^2 u_i}{\partial x_j^2}, \ \ \ \ \ \ \ \ \ \ (1) \]

and

\[ \frac{\partial u_i}{\partial x_i} = 0, \]

in the quarter-circle domain $ D = \{x_1 \geq 0 $, $ x_2 \geq 0 $ and $ x_1^2+x_2^2 \leq 1 \} $, subject to the Dirichlet boundary conditions

\[ \left. \mathbf{u}\right|_{\partial D}=(0,0), \ \ \ \ \ \ \ \ \ \ (2) \]

on the curved and left boundaries; and

\[ \left. \mathbf{u}\right|_{\partial D}=(1,0), \ \ \ \ \ \ \ \ \ \ (3) \]

on the bottom boundary, $ x_2 = 0 $. To make this consistent with Problem 1, we define the body force function as $ (B_1, B_2) = (0,-Re/Fr) $.


Note that in Problem 2, the gravitational body force is represented by the body force rather than the gravity vector.


Switching between the stress-divergence and the simplified forms of the Navier-Stokes equations

The two forms of the Navier-Stokes equations differ in the implementation of the viscous terms, which may be represented as

\[ \frac{\partial^2 u_i}{\partial x_j^2} \mbox{ \ \ \ or \ \ \ } \frac{\partial }{\partial x_j} \left( \frac{\partial u_i}{\partial x_j} + \frac{\partial u_j}{\partial x_i} \right). \]

For an incompressible flow, $ \partial u_i/\partial x_i = 0$, both forms are mathematically equivalent but the stress-divergence form is required for problems with free surfaces , or for problems in which traction boundary conditions are to be applied.

In order to be able do deal with both cases, oomph-lib's Navier-Stokes elements actually implement the viscous term as

\[ \frac{\partial }{\partial x_j} \left( \frac{\partial u_i}{\partial x_j} + \Gamma_i \, \frac{\partial u_j}{\partial x_i} \right). \]

By default the components of the vector $ \Gamma_i $, are set to 1.0, so that the stress-divergence form is used. The components $ \Gamma_i $ are stored in the static data member

static Vector<double> NavierStokesEquations<DIM>::Gamma

of the NavierStokesEquations<DIM> class which forms the basis for all Navier-Stokes elements in oomph-lib. Its entries are initialised to 1.0. The user may over-write these assignments and thus re-define the values of $ \Gamma $ being used for a specific problem. [In principle, it is possible to use stress-divergence form for the first component of the momentum equations, and the simplified form for the second one, say. However, we do not believe that this is a particularly useful/desirable option and have certainly never used such (slightly bizarre) assignments in any of our own computations.]



Solution to problem 1

The figure below shows "carpet plots" of the velocity and pressure fields as well as a contour plot of the pressure distribution with superimposed streamlines for Problem 1 at a Reynolds number of $ Re=100$ and a ratio of Reynolds and Froude numbers (a measure of gravity on the viscous scale) of $ Re/Fr = 100$. The velocity vanishes along the entire domain boundary, apart from the bottom boundary $ (x_2 = 0) $ where the moving "lid" imposes a unit tangential velocity which drives a large vortex, centred at $ (x_1,x_2) \approx (0.59,0.22) $. The pressure singularities created by the velocity discontinuities at $ (x_1,x_2)=(0,0) $ and $ (x_1, x_2)=(1,0) $ are well resolved. The pressure plot shows that away from the singularities, the pressure decreases linearly with $ x_2 $, reflecting the effect of the gravitational body forces which acts in the negative $ x_2-$ direction.

Plot of the velocity and pressure fields for problem 1 with Re=100 and Re/Fr=100, computed with adaptive Taylor-Hood elements.

Solution to problem 2

The next figure shows the computational results for Problem 2, obtained from a computation with adaptive Crouzeix-Raviart elements.

Plot of the velocity and pressure fields for problem 2 with Re=100 and Re/Fr=100, computed with adaptive Crouzeix-Raviart elements.


The code

We use a namespace Global_Physical_Variables to define the various parameters: The Reynolds number,

//==start_of_namespace===================================================
/// Namespace for physical parameters
//=======================================================================
{
/// Reynolds number
double Re=100;
Namespace for physical parameters.

the gravity vector $ {\bf G}$, and the ratio of Reynolds and Froude number, $ Re/Fr $, which represents the ratio of gravitational and viscous forces,

/// Reynolds/Froude number
double Re_invFr=100;
/// Gravity vector
Vector<double> Gravity(2);
double Re_invFr
Reynolds/Froude number.
Vector< double > Gravity(2)
Gravity vector.

In Problem 2, gravity is introduced via the body force function $ {\bf B}$ which we define such that Problems 1 and 2 are equivalent. (We use the gravity vector $ {\bf G} = (0,-1)$ to specify the direction of gravity, while indicating it magnitude by $ Re/Fr.$)

/// Functional body force
void body_force(const double& time, const Vector<double>& x,
Vector<double>& result)
{
result[0]=0.0;
result[1]=-Re_invFr;
}
void body_force(const double &time, const Vector< double > &x, Vector< double > &result)
Functional body force.

Finally we define a body force function, which returns zero values, for use when solving Problem 1.

/// Zero functional body force
void zero_body_force(const double& time, const Vector<double>& x,
Vector<double>& result)
{
result[0]=0.0;
result[1]=0.0;
}
} // end_of_namespace
void zero_body_force(const double &time, const Vector< double > &x, Vector< double > &result)
Zero functional body force.


The driver code

First we create a DocInfo object to control the output, and set the maximum number of spatial adaptations to three.

//==start_of_main======================================================
/// Driver for QuarterCircleDrivenCavityProblem test problem
//=====================================================================
int main()
{
// Set output directory and initialise count
DocInfo doc_info;
doc_info.set_directory("RESLT");
// Set max. number of black-box adaptation
unsigned max_adapt=3;
int main()
Driver for QuarterCircleDrivenCavityProblem test problem.

To solve problem 1 we define the direction of gravity, $ {\bf G} = (0,-1)$, and set the entries in the NavierStokesEquations<2>::Gamma vector to $ (1,1)$, so that the stress-divergence form of the equation is used [In fact, this step is not strictly necessary as it simply re-assigns the default values.]

// Solve problem 1 with Taylor-Hood elements
//--------------------------------------------
{
// Set up downwards-Gravity vector
// Set up Gamma vector for stress-divergence form
NavierStokesEquations<2>::Gamma[0]=1;
NavierStokesEquations<2>::Gamma[1]=1;

Next we build problem 1 using Taylor-Hood elements and passing a function pointer to the zero_body_force(...) function (defined in the namespace Global_Physical_Variables) as the argument.

// Build problem with Gravity vector in stress divergence form,
// using zero body force function
/////////////////////////////////////////////////////////////////// /////////////////////////////////...

Now problem 1 can be solved as in the previous example.

// Solve the problem with automatic adaptation
problem.newton_solve(max_adapt);
// Step number
doc_info.number()=0;
// Output solution
problem.doc_solution(doc_info);
} // end of problem 1

To solve problem 2 we set the entries in the NavierStokesEquations<2>::Gamma vector to zero (thus choosing the simplified version of the Navier-Stokes equations), define $ {\bf G} = (0,0)$, and pass a function pointer to the body_force(...) function to the problem constructor.

// Solve problem 2 with Taylor Hood elements
//--------------------------------------------
{
// Set up zero-Gravity vector
// Set up Gamma vector for simplified form
NavierStokesEquations<2>::Gamma[0]=0;
NavierStokesEquations<2>::Gamma[1]=0;
// Build problem with body force function and simplified form,
// using body force function

Problem 2 may then be solved as before.

// Solve the problem with automatic adaptation
problem.newton_solve(max_adapt);
// Step number
doc_info.number()=1;
// Output solution
problem.doc_solution(doc_info);
} // end of problem 2
} // end_of_main


The problem class

The problem class is very similar to that used in the previous example , with two exceptions:

  • We pass a function pointer to the body force function $ {\bf B}$ to the constructor and
  • store the function pointer to the body force function in the problem's private member data.
//==start_of_problem_class============================================
/// Driven cavity problem in quarter circle domain, templated
/// by element type.
//====================================================================
template<class ELEMENT>
class QuarterCircleDrivenCavityProblem : public Problem
{
public:
/// Constructor
NavierStokesEquations<2>::NavierStokesBodyForceFctPt body_force_fct_pt);
/// Destructor: Empty
/// Update the after solve (empty)
/// Update the problem specs before solve.
/// (Re-)set velocity boundary conditions just to be on the safe side...
{
// Setup tangential flow along boundary 0:
unsigned ibound=0;
unsigned num_nod= mesh_pt()->nboundary_node(ibound);
for (unsigned inod=0;inod<num_nod;inod++)
{
// Tangential flow
unsigned i=0;
mesh_pt()->boundary_node_pt(ibound,inod)->set_value(i,1.0);
// No penetration
i=1;
mesh_pt()->boundary_node_pt(ibound,inod)->set_value(i,0.0);
}
// Overwrite with no flow along all other boundaries
unsigned num_bound = mesh_pt()->nboundary();
for(unsigned ibound=1;ibound<num_bound;ibound++)
{
unsigned num_nod= mesh_pt()->nboundary_node(ibound);
for (unsigned inod=0;inod<num_nod;inod++)
{
for (unsigned i=0;i<2;i++)
{
mesh_pt()->boundary_node_pt(ibound,inod)->set_value(i,0.0);
}
}
}
} // end_of_actions_before_newton_solve
/// After adaptation: Unpin pressure and pin redudant pressure dofs.
{
// Unpin all pressure dofs
RefineableNavierStokesEquations<2>::
unpin_all_pressure_dofs(mesh_pt()->element_pt());
// Pin redundant pressure dofs
RefineableNavierStokesEquations<2>::
pin_redundant_nodal_pressures(mesh_pt()->element_pt());
// Now pin the first pressure dof in the first element and set it to 0.0
fix_pressure(0,0,0.0);
} // end_of_actions_after_adapt
/// Doc the solution
void doc_solution(DocInfo& doc_info);
private:
/// Pointer to body force function
NavierStokesEquations<2>::NavierStokesBodyForceFctPt Body_force_fct_pt;
/// Fix pressure in element e at pressure dof pdof and set to pvalue
void fix_pressure(const unsigned &e, const unsigned &pdof,
const double &pvalue)
{
//Cast to proper element and fix pressure
dynamic_cast<ELEMENT*>(mesh_pt()->element_pt(e))->
fix_pressure(pdof,pvalue);
} // end_of_fix_pressure
}; // end_of_problem_class
~QuarterCircleDrivenCavityProblem()
Destructor: Empty.
void fix_pressure(const unsigned &e, const unsigned &pdof, const double &pvalue)
Fix pressure in element e at pressure dof pdof and set to pvalue.
void doc_solution(DocInfo &doc_info)
Doc the solution.
void actions_after_adapt()
After adaptation: Unpin pressure and pin redudant pressure dofs.
NavierStokesEquations< 2 >::NavierStokesBodyForceFctPt Body_force_fct_pt
Pointer to body force function.
void actions_before_newton_solve()
Update the problem specs before solve. (Re-)set velocity boundary conditions just to be on the safe s...
void actions_after_newton_solve()
Update the after solve (empty)
QuarterCircleDrivenCavityProblem(NavierStokesEquations< 2 >::NavierStokesBodyForceFctPt body_force_fct_pt)
Constructor.


The problem constructor

We store the function pointer to the body force function in the private data member Body_force_fct_pt.

//==start_of_constructor==================================================
/// Constructor for driven cavity problem in quarter circle domain
//========================================================================
template<class ELEMENT>
NavierStokesEquations<2>::NavierStokesBodyForceFctPt body_force_fct_pt) :
Body_force_fct_pt(body_force_fct_pt)

As usual the first task is to create the mesh. We now use the RefineableQuarterCircleSectorMesh<ELEMENT>, which requires the creation of a GeomObject to describe geometry of the curved wall: We choose an ellipse with unit half axes (i.e. a unit circle).

{
// Build geometric object that parametrises the curved boundary
// of the domain
// Half axes for ellipse
double a_ellipse=1.0;
double b_ellipse=1.0;
// Setup elliptical ring
GeomObject* Wall_pt=new Ellipse(a_ellipse,b_ellipse);
// End points for wall
double xi_lo=0.0;
double xi_hi=2.0*atan(1.0);
//Now create the mesh
double fract_mid=0.5;
Problem::mesh_pt() = new
RefineableQuarterCircleSectorMesh<ELEMENT>(
Wall_pt,xi_lo,fract_mid,xi_hi);

Next the error estimator is set, the boundary nodes are pinned and the Reynolds number is assigned, as before .

// Set error estimator
Z2ErrorEstimator* error_estimator_pt=new Z2ErrorEstimator;
dynamic_cast<RefineableQuarterCircleSectorMesh<ELEMENT>*>(
mesh_pt())->spatial_error_estimator_pt()=error_estimator_pt;
// Set the boundary conditions for this problem: All nodes are
// free by default -- just pin the ones that have Dirichlet conditions
// here: All boundaries are Dirichlet boundaries.
unsigned num_bound = mesh_pt()->nboundary();
for(unsigned ibound=0;ibound<num_bound;ibound++)
{
unsigned num_nod= mesh_pt()->nboundary_node(ibound);
for (unsigned inod=0;inod<num_nod;inod++)
{
// Loop over values (u and v velocities)
for (unsigned i=0;i<2;i++)
{
mesh_pt()->boundary_node_pt(ibound,inod)->pin(i);
}
}
} // end loop over boundaries
//Find number of elements in mesh
unsigned n_element = mesh_pt()->nelement();
// Loop over the elements to set up element-specific
// things that cannot be handled by constructor: Pass pointer to Reynolds
// number
for(unsigned e=0;e<n_element;e++)
{
// Upcast from GeneralisedElement to the present element
ELEMENT* el_pt = dynamic_cast<ELEMENT*>(mesh_pt()->element_pt(e));
//Set the Reynolds number, etc
el_pt->re_pt() = &Global_Physical_Variables::Re;

Within this loop we also pass the pointers to $ Re/Fr $, the gravity vector and the body-force function to the elements.

//Set the Re/Fr
el_pt->re_invfr_pt() = &Global_Physical_Variables::Re_invFr;
//Set Gravity vector
//set body force function
el_pt->body_force_fct_pt() = Body_force_fct_pt;
} // end loop over elements

The RefineableQuarterCircleSectorMesh<ELEMENT> contains only three elements and therefore provides a very coarse discretisation of the domain. We refine the mesh uniformly twice before pinning the redundant pressure degrees of freedom, pinning a single pressure degree of freedom, and assigning the equation numbers, as before.

// Initial refinement level
refine_uniformly();
refine_uniformly();
// Pin redudant pressure dofs
RefineableNavierStokesEquations<2>::
pin_redundant_nodal_pressures(mesh_pt()->element_pt());
// Now pin the first pressure dof in the first element and set it to 0.0
fix_pressure(0,0,0.0);
// Setup equation numbering scheme
cout <<"Number of equations: " << assign_eqn_numbers() << std::endl;
} // end_of_constructor


Post processing

The post processing function remains the same as in the previous examples .

//==start_of_doc_solution=================================================
/// Doc the solution
//========================================================================
template<class ELEMENT>
{
ofstream some_file;
char filename[100];
// Number of plot points
unsigned npts=5;
// Output solution
sprintf(filename,"%s/soln%i.dat",doc_info.directory().c_str(),
doc_info.number());
some_file.open(filename);
mesh_pt()->output(some_file,npts);
some_file.close();
} // end_of_doc_solution


Comments and Exercises

  1. Try making the curved boundary the driving wall [Hint: this requires a change in the wall velocities prescribed in Problem::actions_before_newton_solve(). The figure below shows what you should expect.]
Plot of the velocity and pressure distribution for a circular driven cavity in which the flow is driven by the tangential motion of the curvilinear boundary.


Source files for this tutorial



PDF file

A pdf version of this document is available.