Skip to content

mdzik/TCLB_tools

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TCLB tools

Various tools and extended examples for the TCLB project.

Structure

All tools are divided by language: R, Python, etc. In each directory, the subdirectories should be specific packages related to different aspects of interaction with TCLB

Examples and Papers include scripts used to obtain data used in "real world" cases. Be carefoul - they might be computationally intensive. See Readmi for details

Installation

R

Supervisor: Łukasz Łaniewski-Wołłk

devtools::install_github("CFD-GO/TCLB_tools/R/TCLBtools")

Python

Supervisor: Michał Dzikowski

Content of Python/ directory is meant as a packages repository. Add XXXX/Python to your PYTHONPATH than import packages by subdirectory name.

Python: symbolic_tools

Supervisor: Grzegorz Gruszczyński

It is recomended to use PyCharm to run the scripts. It facilitates recognition of paths and folders. Create a PyCharm project by opening the symbolic_tools/ directory in the editor. Run the examples/ . Some are listed below.

Example: Forcing terms

from SymbolicCollisions.core.printers import print_as_vector
from SymbolicCollisions.core.ContinousCMTransforms import \
    ContinousCMTransforms, get_mom_vector_from_continuous_def
from SymbolicCollisions.core.cm_symbols import \
    F3D, dzeta3D, u3D
from SymbolicCollisions.core.cm_symbols import rho, moments_dict


ccmt = ContinousCMTransforms(dzeta3D, u3D, F3D, rho)
lattice = 'D2Q9'

print("\n--- FORCES ---")
print('\n\n// === continuous central moments === \n ')

print('\n//Force -> Force_cm - from continous definition: \n'
      'k_mn = integrate(fun, (x, -oo, oo), (y, -oo, oo)) \n'
      'where fun = forceM(rho,u,x,y) *(x-ux)^m *(y-uy)^n *(z-uz)^o ')
F_cm = get_mom_vector_from_continuous_def(ccmt.get_force_He_MB,
                                          continuous_transformation=ccmt.get_cm,
                                          moments_order=moments_dict[lattice])
print_as_vector(F_cm, 'F_cm')

Output:

//Force -> Force_cm - from continous definition: 
k_mn = integrate(fun, (x, -oo, oo), (y, -oo, oo)) 
where fun = forceM(rho,u,x,y) *(x-ux)^m *(y-uy)^n *(z-uz)^o 
	F_cm[0] = 0;
	F_cm[1] = Fhydro.x*m00/rho;
	F_cm[2] = Fhydro.y*m00/rho;
	F_cm[3] = 0;
	F_cm[4] = 0;
	F_cm[5] = 0;
	F_cm[6] = 1/3.*Fhydro.y*m00/rho;
	F_cm[7] = 1/3.*Fhydro.x*m00/rho;
	F_cm[8] = 0;

Example: Equilibrium Distribution

from SymbolicCollisions.core.printers import print_as_vector
from sympy.matrices import Matrix
from sympy import Symbol
from SymbolicCollisions.core.ContinousCMTransforms import ContinousCMTransforms, get_mom_vector_from_continuous_def
from SymbolicCollisions.core.cm_symbols import \
    F3D, dzeta3D, u3D

from SymbolicCollisions.core.cm_symbols import e_D2Q9, u2D, F2D, rho, moments_dict

from SymbolicCollisions.core.DiscreteCMTransforms import \
    DiscreteCMTransforms, get_mom_vector_from_discrete_def


lattice = 'D2Q9'
ccmt = ContinousCMTransforms(dzeta3D, u3D, F3D, rho)
dcmt = DiscreteCMTransforms(e_D2Q9, u2D, F2D, rho)

print("\n--- EQUILIBRIA ---")
print('// === discrete cm ===\n ')

print('\n//population_eq -> cm_eq - by definition: k_mn = sum( (e_ix-ux)^m (e_iy-uy)^n * population_eq_i)')
print("moments: first order (linear) velocity expansion.")
pop_eq = get_mom_vector_from_discrete_def(lambda i: Symbol('m00') * dcmt.get_gamma_first_order(i),
                                          discrete_transform=dcmt.get_cm,
                                          moments_order=moments_dict[lattice])
print_as_vector(pop_eq, 'cm_eq_first_order')

print("moments: second order (quadratic) velocity expansion.")
pop_eq = get_mom_vector_from_discrete_def(lambda i: Symbol('m00') * dcmt.get_gamma(i),
                                          discrete_transform=dcmt.get_cm,
                                          moments_order=moments_dict[lattice])
print_as_vector(pop_eq, 'cm_eq_second_order')

print('\n\n// === continous cm === \n ')
# to calculate particular moment
row = moments_dict['D2Q9'][0]
moment = ccmt.get_cm(row, ccmt.get_Maxwellian_DF)
print_as_vector(Matrix([moment]), 'particular_moment')


print('\n//population_eq -> cm_eq - from continous definition: \n'
      'k_mn = integrate(fun, (x, -oo, oo), (y, -oo, oo)) \n'
      'where fun = fM(rho,u,x,y) *(x-ux)^m *(y-uy)^n *(z-uz)^o ')
cm_eq = get_mom_vector_from_continuous_def(ccmt.get_Maxwellian_DF,
                                           continuous_transformation=ccmt.get_cm,
                                           moments_order=moments_dict[lattice])
print_as_vector(cm_eq, 'cm_eq')

Output

--- EQUILIBRIA ---
// === discrete cm ===
 

//population_eq -> cm_eq - by definition: k_mn = sum( (e_ix-ux)^m (e_iy-uy)^n * population_eq_i)
moments: first order (linear) velocity expansion.
	cm_eq_first_order[0] = m00;
	cm_eq_first_order[1] = 0;
	cm_eq_first_order[2] = 0;
	cm_eq_first_order[3] = m00*(-ux2 + 1/3.);
	cm_eq_first_order[4] = m00*(-uy2 + 1/3.);
	cm_eq_first_order[5] = -m00*uxuy;
	cm_eq_first_order[6] = 2.*m00*ux2*u.y;
	cm_eq_first_order[7] = 2.*m00*u.x*uy2;
	cm_eq_first_order[8] = m00*(-3.*ux2*uy2 - 1/3.*ux2 - 1/3.*uy2 + 1/9.);
moments: second order (quadratic) velocity expansion.
	cm_eq_second_order[0] = m00;
	cm_eq_second_order[1] = 0;
	cm_eq_second_order[2] = 0;
	cm_eq_second_order[3] = 1/3.*m00;
	cm_eq_second_order[4] = 1/3.*m00;
	cm_eq_second_order[5] = 0;
	cm_eq_second_order[6] = -m00*ux2*u.y;
	cm_eq_second_order[7] = -m00*u.x*uy2;
	cm_eq_second_order[8] = m00*(3.*ux2*uy2 + 1/9.);


// === continous cm === 
 
	particular_moment[0] = m00;

//population_eq -> cm_eq - from continous definition: 
k_mn = integrate(fun, (x, -oo, oo), (y, -oo, oo)) 
where fun = fM(rho,u,x,y) *(x-ux)^m *(y-uy)^n *(z-uz)^o 
	cm_eq[0] = m00;
	cm_eq[1] = 0;
	cm_eq[2] = 0;
	cm_eq[3] = 1/3.*m00;
	cm_eq[4] = 1/3.*m00;
	cm_eq[5] = 0;
	cm_eq[6] = 0;
	cm_eq[7] = 0;
	cm_eq[8] = 1/9.*m00;

Example: Collision Kernel

from sympy.matrices import eye
from sympy.printing import print_ccode
from SymbolicCollisions.core.cm_symbols import omega_ade, omega_b, omega_v, m00
from SymbolicCollisions.core.cm_symbols import dynamic_import
from SymbolicCollisions.core.DiscreteCMTransforms import get_DF, get_m00
from SymbolicCollisions.core.printers import print_u2, print_as_vector
from SymbolicCollisions.core.MatrixGenerator import get_raw_moments_matrix, get_shift_matrix

# inspired by:
# "Consistent Forcing Scheme in the cascaded LBM" L. Fei et al. 2017
# eqs 8-12 : (eye(q)-S)*cm + S*cm_eq + (eye(q)-S/2.)*force_in_cm_space

# SETUP
d = 2
q = 9
model = 'hydro'  # choose from '['hydro', 'ade', 'ade_with_f']

# DYNAMIC IMPORTS
ex = dynamic_import("SymbolicCollisions.core.cm_symbols", f"ex_D{d}Q{q}")
ey = dynamic_import("SymbolicCollisions.core.cm_symbols", f"ey_D{d}Q{q}")
if d == 3:
    ez = dynamic_import("SymbolicCollisions.core.cm_symbols", f"ez_D{d}Q{q}")
else:
    ez = None


def get_s_relax_switcher(choice):
    s_relax_switcher = {
        'hydro': ("SymbolicCollisions.core.cm_symbols", f"S_relax_hydro_D{d}Q{q}"),
        'ade_with_f': ("SymbolicCollisions.core.cm_symbols", f"S_relax_ADE_D{d}Q{q}"),
        'ade': ("SymbolicCollisions.core.cm_symbols", f"S_relax_ADE_D{d}Q{q}"),
    }
    which_model = s_relax_switcher.get(choice, lambda: "Invalid argument")
    return dynamic_import(*which_model)


S_Relax = get_s_relax_switcher(model)

hardcoded_cm_eq = dynamic_import("SymbolicCollisions.core.hardcoded_results", f"hardcoded_cm_eq_compressible_D{d}Q{q}")
hardcoded_F_cm = dynamic_import("SymbolicCollisions.core.hardcoded_results", f"hardcoded_F_cm_pf_D{d}Q{q}")
from SymbolicCollisions.core.cm_symbols import Force_str as F_str

# ARRANGE STUFF
Mraw = get_raw_moments_matrix(ex, ey, ez)
Nraw = get_shift_matrix(Mraw.inv(), ex, ey, ez)

# from sympy import pprint
# pprint(Mraw)  # see what you have done
# pprint(Nraw)

pop_in_str = 'x_in'  # symbol defining populations
temp_pop_str = 'temp'  # symbol defining populations
cm_eq_pop_str = 'cm_eq'  # symbol defining populations


# GENERATE CODE
def make_header(choice):
    model_switcher = {
        'hydro': f"CudaDeviceFunction void relax_and_collide_hydro_with_F(real_t {pop_in_str}[{q}], real_t {omega_v}, vector_t u, vector_t {F_str}) \n{{",
        'ade_with_f': f"CudaDeviceFunction void relax_and_collide_ADE_with_F(real_t {pop_in_str}[{q}], real_t {omega_ade}, vector_t u, vector_t {F_str}) \n{{",
        'ade': f"CudaDeviceFunction void relax_and_collide_ADE(real_t {pop_in_str}[{q}], real_t {omega_ade}, vector_t u) \n{{",
    }
    result = model_switcher.get(choice, lambda: "Invalid argument")
    print(result)


make_header(model)

print("\t//=== THIS IS AUTOMATICALLY GENERATED CODE ===")
# print(f"real_t {sv} = omega;")
# print("real_t bulk_visc = 1./6. ;")
# print("real_t {sb} = 1./(3*bulk_visc + 0.5);")
# print(f"real_t {sb} = omega_bulk;\n")  # s_b = 1./(3*bulk_visc + 0.5)
print_u2(d)
print_ccode(get_m00(q, pop_in_str), assign_to=f'\treal_t {m00}')


def make_variables(choice):
    model_switcher = {
        'hydro': f"\n\treal_t {temp_pop_str}[{q}];\n",
        'ade_with_f': f"\n\treal_t {temp_pop_str}[{q}];\n",
        'ade': f"\n\treal_t {temp_pop_str}[{q}];\n",
    }
    # Get the function from switcher dictionary
    result = model_switcher.get(choice, lambda: "Invalid argument")
    print(result)


make_variables(model)

print(f"\tfor (int i = 0; i < {q}; i++) {{\n\t"
      f"\t{temp_pop_str}[i] = {pop_in_str}[i];}}")

populations = get_DF(q, pop_in_str)
temp_populations = get_DF(q, temp_pop_str)
cm_eq = get_DF(q, cm_eq_pop_str)
F_cm = get_DF(q, F_str)
m = Mraw * temp_populations

print("\n\t//raw moments from density-probability functions")
# print("\t//[m00, m10, m01, m20, m02, m11, m21, m12, m22]")
print_as_vector(m, print_symbol=pop_in_str)

print("\n\t//central moments from raw moments")
cm = Nraw * populations
print_as_vector(cm, print_symbol=temp_pop_str)

print("\n\t//collision in central moments space")
# print("//calculate equilibrium distributions in cm space")
# print("real_t {cm_eq_pop_str}[{q}];\n")
# print_as_vector(hardcoded_cm_eq, cm_eq_pop_str)  # save time, verbosity
# print("//calculate forces in cm space")
# print("real_t {F_cm_str}[{q}];")
# print_as_vector(hardcoded_F_cm, F_cm_str)  # save time, verbosity
print("\t//collide")


def make_collision(choice):
    model_switcher = {
        # Relax 2nd moments for hydro, SOI
        'hydro': (eye(q) - S_Relax) * temp_populations
                 + S_Relax * hardcoded_cm_eq
                 + (eye(q) - S_Relax / 2) * hardcoded_F_cm,

        # Relax 1st moments for ADE, SOI
        'ade_with_f': (eye(q) - S_Relax) * temp_populations
                      + S_Relax * hardcoded_cm_eq
                      + (eye(q) - S_Relax / 2) * hardcoded_F_cm,
        # Relax 1st moments for ADE, SOI without force
        'ade': (eye(q) - S_Relax) * temp_populations
               + S_Relax * hardcoded_cm_eq,
    }
    # Get the function from switcher dictionary
    cm_after_collision = model_switcher.get(choice, lambda: "Invalid argument")
    print_as_vector(cm_after_collision, print_symbol=pop_in_str)


make_collision(model)
print("\n\t//back to raw moments")
m = Nraw.inv() * populations
print_as_vector(m, print_symbol=temp_pop_str)

print("\n\t//back to density-probability functions")
populations = Mraw.inv() * temp_populations
print_as_vector(populations, print_symbol=pop_in_str)

print("\n}\n")

Output

CudaDeviceFunction void relax_and_collide_hydro_with_F(real_t x_in[9], real_t omega_nu, vector_t u, vector_t F) 
{
	//=== THIS IS AUTOMATICALLY GENERATED CODE ===
	real_t uxuy = u.x*u.y;
	real_t ux2 = u.x*u.x;
	real_t uy2 = u.y*u.y;

real_t m00 = x_in[0] + x_in[1] + x_in[2] + x_in[3] + x_in[4] + x_in[5] + x_in[6] + x_in[7] + x_in[8];

	real_t temp[9];

	for (int i = 0; i < 9; i++) {
		temp[i] = x_in[i];}

	//raw moments from density-probability functions
	x_in[0] = temp[0] + temp[1] + temp[2] + temp[3] + temp[4] + temp[5] + temp[6] + temp[7] + temp[8];
	x_in[1] = temp[1] - temp[3] + temp[5] - temp[6] - temp[7] + temp[8];
	x_in[2] = temp[2] - temp[4] + temp[5] + temp[6] - temp[7] - temp[8];
	x_in[3] = temp[1] + temp[3] + temp[5] + temp[6] + temp[7] + temp[8];
	x_in[4] = temp[2] + temp[4] + temp[5] + temp[6] + temp[7] + temp[8];
	x_in[5] = temp[5] - temp[6] + temp[7] - temp[8];
	x_in[6] = temp[5] + temp[6] - temp[7] - temp[8];
	x_in[7] = temp[5] - temp[6] - temp[7] + temp[8];
	x_in[8] = temp[5] + temp[6] + temp[7] + temp[8];

	//central moments from raw moments
	temp[0] = x_in[0];
	temp[1] = -u.x*x_in[0] + x_in[1];
	temp[2] = -u.y*x_in[0] + x_in[2];
	temp[3] = ux2*x_in[0] - 2.*u.x*x_in[1] + x_in[3];
	temp[4] = uy2*x_in[0] - 2.*u.y*x_in[2] + x_in[4];
	temp[5] = uxuy*x_in[0] - u.x*x_in[2] - u.y*x_in[1] + x_in[5];
	temp[6] = -ux2*u.y*x_in[0] + ux2*x_in[2] + 2.*uxuy*x_in[1] - 2.*u.x*x_in[5] - u.y*x_in[3] + x_in[6];
	temp[7] = -u.x*uy2*x_in[0] + 2.*uxuy*x_in[2] - u.x*x_in[4] + uy2*x_in[1] - 2.*u.y*x_in[5] + x_in[7];
	temp[8] = ux2*uy2*x_in[0] - 2.*ux2*u.y*x_in[2] + ux2*x_in[4] - 2.*u.x*uy2*x_in[1] + 4.*uxuy*x_in[5] - 2.*u.x*x_in[7] + uy2*x_in[3] - 2.*u.y*x_in[6] + x_in[8];

	//collision in central moments space
	//collide
	x_in[0] = m00;
	x_in[1] = 1/2.*F.x;
	x_in[2] = 1/2.*F.y;
	x_in[3] = 1/3.*m00*omega_bulk - 1/2.*omega_bulk*temp[3] - 1/2.*omega_bulk*temp[4] - 1/2.*omega_nu*temp[3] + 1/2.*omega_nu*temp[4] + temp[3];
	x_in[4] = 1/3.*m00*omega_bulk - 1/2.*omega_bulk*temp[3] - 1/2.*omega_bulk*temp[4] + 1/2.*omega_nu*temp[3] - 1/2.*omega_nu*temp[4] + temp[4];
	x_in[5] = -temp[5]*(omega_nu - 1.);
	x_in[6] = 1/6.*F.y;
	x_in[7] = 1/6.*F.x;
	x_in[8] = 1/9.*m00;

	//back to raw moments
	temp[0] = x_in[0];
	temp[1] = u.x*x_in[0] + x_in[1];
	temp[2] = u.y*x_in[0] + x_in[2];
	temp[3] = ux2*x_in[0] + 2.*u.x*x_in[1] + x_in[3];
	temp[4] = uy2*x_in[0] + 2.*u.y*x_in[2] + x_in[4];
	temp[5] = uxuy*x_in[0] + u.x*x_in[2] + u.y*x_in[1] + x_in[5];
	temp[6] = ux2*u.y*x_in[0] + ux2*x_in[2] + 2.*uxuy*x_in[1] + 2.*u.x*x_in[5] + u.y*x_in[3] + x_in[6];
	temp[7] = u.x*uy2*x_in[0] + 2.*uxuy*x_in[2] + u.x*x_in[4] + uy2*x_in[1] + 2.*u.y*x_in[5] + x_in[7];
	temp[8] = ux2*uy2*x_in[0] + 2.*ux2*u.y*x_in[2] + ux2*x_in[4] + 2.*u.x*uy2*x_in[1] + 4.*uxuy*x_in[5] + 2.*u.x*x_in[7] + uy2*x_in[3] + 2.*u.y*x_in[6] + x_in[8];

	//back to density-probability functions
	x_in[0] = temp[0] - temp[3] - temp[4] + temp[8];
	x_in[1] = 1/2.*temp[1] + 1/2.*temp[3] - 1/2.*temp[7] - 1/2.*temp[8];
	x_in[2] = 1/2.*temp[2] + 1/2.*temp[4] - 1/2.*temp[6] - 1/2.*temp[8];
	x_in[3] = -1/2.*temp[1] + 1/2.*temp[3] + 1/2.*temp[7] - 1/2.*temp[8];
	x_in[4] = -1/2.*temp[2] + 1/2.*temp[4] + 1/2.*temp[6] - 1/2.*temp[8];
	x_in[5] = 1/4.*temp[5] + 1/4.*temp[6] + 1/4.*temp[7] + 1/4.*temp[8];
	x_in[6] = -1/4.*temp[5] + 1/4.*temp[6] - 1/4.*temp[7] + 1/4.*temp[8];
	x_in[7] = 1/4.*temp[5] - 1/4.*temp[6] - 1/4.*temp[7] + 1/4.*temp[8];
	x_in[8] = -1/4.*temp[5] - 1/4.*temp[6] + 1/4.*temp[7] + 1/4.*temp[8];

}

About

Various tools for the TCLB project

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 61.5%
  • Jupyter Notebook 35.9%
  • C++ 1.7%
  • MATLAB 0.6%
  • M4 0.2%
  • R 0.1%