Elastodynamics + Contact Mechanics

I am working on a test script (FEniCSx-real in docker) that combines elastodynamics (modified from Elastodynamics Example ) with a penalty method to prevent penetration of a rigid surface. In the test (the MWE included below), an elastic cube, starting at rest, is in free fall in a gravitational field over a rigid surface. The elastodynamics equations are iterated to solve the displacement of the cube, and the penalty method used at each step to prevent penetration of the rigid surface using the absolute position of the cube (displacement + reference configuration).

However, right now, if there is no initial penetration, then the object just remains in freefall and the center of mass (My in the code) follows a parabola. If there is some initial penetration, then the object follows an upward parabola (constant upward force). From this, it seems like the penalty term is staying at its initial value and not updating in the solver. If I print out values during the iterations, the values are changing.

Are there particular ways that penalty methods have to be set up in FEniCSx if the level of penetration is a function of absolute position as opposed to the displacement (as is done in the Hertzian contact example)?

The MWE of the elastodynamics/contact problem is included below:

import numpy as np
import ufl
import pyvista
import matplotlib.pyplot as plt

from petsc4py import PETSc
from mpi4py import MPI
from dolfinx import fem, mesh, plot
from dolfinx import *

#### Defining domain and function spaces
domain = mesh.create_rectangle(MPI.COMM_WORLD,[[0.0,0.0],[1,1]],[100,10],mesh.CellType.triangle)
V = fem.VectorFunctionSpace(domain,("CG",2))    # function space for the displacements and test functions
Vsig = fem.TensorFunctionSpace(domain,("DG",0)) # function space for the stress tensor

# gravity vector
b = -0.1

B = fem.Constant(domain,PETSc.ScalarType((0,b)))

#### Getting facets of the bottom edge that will come in contact ####
def bottom(x):
    return np.isclose(x[1],0)

fdim = domain.topology.dim -1 
bottom_facets = mesh.locate_entities_boundary(domain,fdim,bottom)

marked_facets = np.hstack([bottom_facets])
marked_values = np.hstack([np.full(len(bottom_facets),3,dtype=np.int32)])
sorted_facets = np.argsort(marked_facets)
facet_tag = mesh.meshtags(domain,fdim,marked_facets[sorted_facets],marked_values[sorted_facets])


#### Defining elasticity constants ####
E = 1000.0
nu = 0.3
mu = fem.Constant(domain,E / (2.0*(1+nu)))
lmbda = fem.Constant(domain,E*nu / ((1.0+nu)*(1.0-2.0*nu)))

k_pen = fem.Constant(domain,100000.0)

rho = 1.0

#### Defining parameters associated with Generalized alpha method and timestepping ####
am = 0.2
af = 0.4
g = 0.5 + af - am

alpha_m = fem.Constant(domain,am)
alpha_f = fem.Constant(domain,af)
gamma = fem.Constant(domain,g)
beta = fem.Constant(domain,(g+0.5)**2 / 4.0)

T = 10
Nsteps = 100
dt = fem.Constant(domain,T/Nsteps)

#### Defining functions of interest ####
u_ = ufl.TestFunction(V)
u = fem.Function(V,name='Displacement')

u_old = fem.Function(V)
v_old = fem.Function(V)
a_old = fem.Function(V)

penetrate = fem.Function(V) # value of the penetration at each element
abs_penetrate = fem.Function(V)

X0 = fem.Function(V) # Reference Configuration of the domain
X0.interpolate(lambda x: [x[0],x[1]])
X = fem.Function(V) # Current Configuration of the domain
X.interpolate(X0)

NN = len(X.x.array)
# array indices of the x- and y-components of the vector-valued function arrays
x_inds = np.arange(0,NN,2)
y_inds = np.arange(1,NN+1,2)


#### Defining measures ####
metadata = {"quadrature_degree":4}
ds = ufl.Measure('ds',domain=domain,subdomain_data=facet_tag,metadata=metadata)
dx = ufl.Measure("dx",domain=domain,metadata=metadata)


#### defining elements of GA method ####
def sigma(r):
    return lmbda*ufl.nabla_div(r)*ufl.Identity(len(r)) + 2*mu*ufl.sym(ufl.grad(r))
    
def m(u,u_):
    return rho*ufl.inner(u,u_)*dx

def k(u,u_):
    return ufl.inner(sigma(u),ufl.grad(u_))*dx

def penalty(u,u_):
    X.x.array[:] = u.x.array[:] + X0.x.array[:]
    penetrate.x.array[y_inds] = 0.0 - X.x.array[y_inds] # Boundary at y=0 
    abs_penetrate.x.array[:] = (penetrate.x.array[:] + np.abs(penetrate.x.array[:]))/2
    return ufl.inner(k_pen*abs_penetrate,u_)*ds(3)

def update_a(u,u_old,v_old,a_old,ufl_=True):
    if ufl_:
        dt_ = dt
        beta_ = beta
    else:
        dt_ = dt
        beta_ = beta.value
        
    return (u-u_old-dt_*v_old)/beta_/dt_**2 - (1-2*beta_)/2/beta_*a_old

def update_v(a,u_old,v_old,a_old,ufl_=True):
    if ufl_:
        dt_ = dt
        gamma_ = gamma
    else:
        dt_ = dt
        gamma_ = gamma.value
    return v_old + dt_*((1-gamma_)*a_old + gamma_*a)

def update_fields(u,u_old,v_old,a_old):
    u_vec,u0_vec = u.x.array[:],u_old.x.array[:]
    v0_vec,a0_vec = v_old.x.array[:],a_old.x.array[:]
    
    a_vec = update_a(u_vec,u0_vec,v0_vec,a0_vec,ufl_=False)
    v_vec = update_v(a_vec,u0_vec,v0_vec,a0_vec,ufl_=False)
    v_old.x.array[:] = v_vec
    a_old.x.array[:] = a_vec
    u_old.x.array[:] = u_vec
    
def avg(x_old,x_new,alpha):
    return alpha*x_old + (1-alpha)*x_new

a_new = update_a(u,u_old,v_old,a_old,ufl_=True)
v_new = update_v(a_new,u_old,v_old,a_old,ufl_=True)

#### Defining the problem and solver ####
res =  m(avg(a_old,a_new,alpha_m),u_) + k(avg(u_old,u,alpha_f),u_) - rho*ufl.inner(B,u_)*dx - penalty(u,u_) #residual to solve
problem = fem.petsc.NonlinearProblem(res,u)#,bcs=bcs)

from dolfinx import nls
solver = nls.petsc.NewtonSolver(domain.comm, problem)

solver.atol = 1e-8
solver.rtol = 1e-8
solver.convergence_criterion = "incremental"

#### Quatities to track ####
time = np.linspace(0, T, Nsteps+1)

V0 = 1

# center of mass positions
Mx_ = np.zeros((Nsteps+1,))
My_ = np.zeros((Nsteps+1,))
Mx = fem.form((1/V0)*(u[0]+X0[0])*dx)
My = fem.form((1/V0)*(u[1]+X0[1])*dx)
Mx_[0] = fem.assemble_scalar(Mx)
My_[0] = fem.assemble_scalar(My)


#### Iterating through time ####
for (i, dt) in enumerate(np.diff(time)):
    
    t = time[i+1]    
    num_its,converged = solver.solve(u) # solve the current time step
    assert(converged)
    u.x.scatter_forward()
    
    # Update old fields with new quantities
    update_fields(u, u_old, v_old, a_old)
    Mx_[i+1] = fem.assemble_scalar(Mx)
    My_[i+1] = fem.assemble_scalar(My)
    
# Plot center of mass evolution
plt.figure()
plt.plot(time,Mx_,time,My_,time,My_[0]+0.5*b*time**2)
plt.legend(("Mx","My","My_freefall"))
plt.xlabel("Time")
plt.ylabel("Center of Mass")
plt.show()   

Thanks so much!

1 Like

I found a working solution where the penalty function is initialized and updated during the “update_fields” function call. I have added the updated MWE with some additional output plots. I also added back in the Rayleigh damping terms used in the original elastodynamics example to help with stability (currently this requires a good bit of tuning, but could probably be improved with adaptive time-stepping).

import numpy as np
import ufl
import pyvista
import matplotlib.pyplot as plt

from petsc4py import PETSc
from mpi4py import MPI
from dolfinx import fem, mesh, plot
from dolfinx import *

#### Defining domain and function spaces
domain = mesh.create_rectangle(MPI.COMM_WORLD,[[0.0,0.0],[1,1]],[10,10],mesh.CellType.triangle)
V = fem.VectorFunctionSpace(domain,("CG",2))    # function space for the displacements and test functions
Vsig = fem.TensorFunctionSpace(domain,("DG",0)) # function space for the stress tensor

# gravity vector
b = -5
K = -0.1 # boundary height

B = fem.Constant(domain,PETSc.ScalarType((0,b)))

#### Getting facets of the bottom edge that will come in contact ####
def bottom(x):
    return np.isclose(x[1],0)

fdim = domain.topology.dim -1 
bottom_facets = mesh.locate_entities_boundary(domain,fdim,bottom)

marked_facets = np.hstack([bottom_facets])
marked_values = np.hstack([np.full(len(bottom_facets),3,dtype=np.int32)])
sorted_facets = np.argsort(marked_facets)
facet_tag = mesh.meshtags(domain,fdim,marked_facets[sorted_facets],marked_values[sorted_facets])


#### Defining elasticity constants ####
E = 100000.0
nu = 0.3
mu = fem.Constant(domain,E / (2.0*(1+nu)))
lmbda = fem.Constant(domain,E*nu / ((1.0+nu)*(1.0-2.0*nu)))

k_pen = 1000000

rho = 1.0
eta_m = 0.01
eta_k = 0.01
#### Defining parameters associated with Generalized alpha method and timestepping ####
am = 0.2
af = 0.4
g = 0.5 + af - am

alpha_m = fem.Constant(domain,am)
alpha_f = fem.Constant(domain,af)
gamma = fem.Constant(domain,g)
beta = fem.Constant(domain,(g+0.5)**2 / 4.0)

T = 3
Nsteps = 10000
dt = fem.Constant(domain,T/Nsteps)

#### Defining functions of interest ####
u_ = ufl.TestFunction(V)
u = fem.Function(V,name='Displacement')

u_old = fem.Function(V)
v_old = fem.Function(V)
a_old = fem.Function(V)


##########################################
###### Independent functions for the #####
###### values related to penetration #####
##########################################
penetrate = fem.Function(V) # value of the penetration at each element
abs_penetrate = fem.Function(V)
penalty_f = fem.Function(V)


X0 = fem.Function(V) # Reference Configuration of the domain
X0.interpolate(lambda x: [x[0],x[1]])
X = fem.Function(V) # Current Configuration of the domain
X.interpolate(X0)

NN = len(X.x.array)
# array indices of the x- and y-components of the vector-valued function arrays
x_inds = np.arange(0,NN,2)
y_inds = np.arange(1,NN+1,2)

#v_old.x.array[x_inds] = 5
#v_old.x.array[y_inds] = 5

#### Defining measures ####
metadata = {"quadrature_degree":4}
ds = ufl.Measure('ds',domain=domain,subdomain_data=facet_tag,metadata=metadata)
dx = ufl.Measure("dx",domain=domain,metadata=metadata)


#### defining elements of GA method ####
def sigma(r):
    return lmbda*ufl.nabla_div(r)*ufl.Identity(len(r)) + 2*mu*ufl.sym(ufl.grad(r))
    
def m(u,u_):
    return rho*ufl.inner(u,u_)*dx

def k(u,u_):
    return ufl.inner(sigma(u),ufl.grad(u_))*dx

def c(u,u_):
    return eta_m*m(u,u_) + eta_k*k(u,u_)


##########################################
##### Independent function to update #####
##### the penalty function, used in  #####
##### in the update_fields function. #####
##########################################

def update_penalty(u,X,X0,penetrate):
    X.x.array[:] = u.x.array + X0.x.array
    penetrate.x.array[y_inds] = K - X.x.array[y_inds] # Boundary at y=K 
    penetrate.x.array[x_inds] = 0
    return 0.5*k_pen*(penetrate.x.array[:] + np.abs(penetrate.x.array[:]))

##########################################
##### Calculates the work associated #####
##### with the penalty as a Form to  #####
##### incorporate with the residual. #####
##########################################

def penalty(penalty_f,u_):
    return ufl.inner(penalty_f,u_)*ds


def update_a(u, u_old, v_old, a_old, ufl_=True):
    if ufl_:
        dt_ = dt
        beta_ = beta
    else:
        dt_ = dt
        beta_ = beta.value
    return (u-u_old-dt_*v_old)/beta_/dt_**2 - (1-2*beta_)/2/beta_*a_old


def update_v(a,u_old,v_old,a_old,ufl_=True):
    if ufl_:
        dt_ = dt
        gamma_ = gamma
    else:
        dt_ = dt
        gamma_ = gamma.value
    return v_old + dt_*((1-gamma_)*a_old + gamma_*a)

def update_fields(u,u_old,v_old,a_old):
    u_vec,u0_vec = u.x.array[:],u_old.x.array[:]
    v0_vec,a0_vec = v_old.x.array[:],a_old.x.array[:]
    
    a_vec = update_a(u_vec,u0_vec,v0_vec,a0_vec,ufl_=False)
    v_vec = update_v(a_vec,u0_vec,v0_vec,a0_vec,ufl_=False)
    v_old.x.array[:] = v_vec
    a_old.x.array[:] = a_vec
    u_old.x.array[:] = u_vec
    ########################################
    ##### Update the penalty function. #####
    ########################################
    penalty_f.x.array[:] = update_penalty(u,X,X0,penetrate)
    
def avg(x_old,x_new,alpha):
    return alpha*x_old + (1-alpha)*x_new

a_new = update_a(u,u_old,v_old,a_old,ufl_=True)
v_new = update_v(a_new,u_old,v_old,a_old,ufl_=True)

##########################################
##### Initialize the penalty function ####
##########################################
penalty_f.x.array[:] = update_penalty(u,X,X0,penetrate)

#### Defining the problem and solver ####
res =  m(avg(a_old,a_new,alpha_m),u_) + c(avg(v_old,v_new,alpha_f),u_) + k(avg(u_old,u,alpha_f),u_) - rho*ufl.inner(B,u_)*dx - penalty(penalty_f,u_) #residual to solve
problem = fem.petsc.NonlinearProblem(res,u)#,bcs=bcs)

from dolfinx import nls
solver = nls.petsc.NewtonSolver(domain.comm, problem)

solver.atol = 1e-8
solver.rtol = 1e-8
solver.convergence_criterion = "incremental"

#### Quatities to track ####
time = np.linspace(0, T, Nsteps+1)

V0 = 1

# center of mass positions
Mx_ = np.zeros((Nsteps+1,))
My_ = np.zeros((Nsteps+1,))
Mx = fem.form((1/V0)*(u[0]+X0[0])*dx)
My = fem.form((1/V0)*(u[1]+X0[1])*dx)
Mx_[0] = fem.assemble_scalar(Mx)
My_[0] = fem.assemble_scalar(My)
max_pen = np.zeros((Nsteps+1,))

from dolfinx import log
#log.set_log_level(log.LogLevel.INFO)
#### Iterating through time ####
for (i, dt) in enumerate(np.diff(time)):
    t = time[i+1]   
    num_its,converged = solver.solve(u) # solve the current time step
    assert(converged)
    u.x.scatter_forward()
    
    # Update old fields with new quantities
    update_fields(u, u_old, v_old, a_old)
        
    Mx_[i+1] = fem.assemble_scalar(Mx)
    My_[i+1] = fem.assemble_scalar(My)
    max_pen[i+1] = max(penalty_f.x.array[y_inds])/k_pen ## Maximum penetration of the cube

# Plot center of mass evolution
plt.figure()
plt.plot(time,Mx_,time,My_)
plt.legend(("Mx","My"))
plt.xlabel("Time")
plt.ylabel("Center of Mass")
plt.show()   

# Plot of the maximum penetration
plt.figure()
plt.plot(time,max_pen)
plt.legend(("Max Penetration"))
plt.xlabel("Time")
plt.ylabel("Penetration")
plt.show()
1 Like