Develop Reference

How to simulate a heat diffusion on a rectangular ring with FiPy?

I am new to solving a PDE and experimenting with a heat diffusion on a copper body of a rectangular ring shape using FiPy.
And this is a plot of simulation result at some times.
I am using the Grid2D() for a mesh and the CellVariable.constrain() to specify boundary conditions. The green dots are centers of exterior faces where T = 273.15 + 25 (K), and blue dots are centers of interior faces where T = 273.15 + 30 (K).
Obviously, I am doing something wrong, because the temperature goes down to 0K. How should I specify boundary conditions correctly?
These are the code.
import numpy as np
import matplotlib.pyplot as plt
import fipy
def get_mask_of_rect(mesh, x, y, w, h):
def left_id(i, j): return mesh.numberOfHorizontalFaces + i*mesh.numberOfVerticalColumns + j
def right_id(i, j): return mesh.numberOfHorizontalFaces + i*mesh.numberOfVerticalColumns + j + 1
def bottom_id(i, j): return i*mesh.nx + j
def top_id(i, j): return (i+1)*mesh.nx + j
j0, i0 = np.floor(np.array([x, y]) / [mesh.dx, mesh.dy]).astype(int)
n, m = np.round(np.array([w, h]) / [mesh.dx, mesh.dy]).astype(int)
mask = np.zeros_like(mesh.exteriorFaces, dtype=bool)
for i in range(i0, i0 + n):
mask[left_id(i, j0)] = mask[right_id(i, j0 + m-1)] = True
for j in range(j0, j0 + m):
mask[bottom_id(i0, j)] = mask[top_id(i0 + n-1, j)] = True
return mask
mesh = fipy.Grid2D(Lx = 1, Ly = 1, nx = 20, ny = 20) # Grid of size 1m x 1m
k_over_c_rho = 3.98E2 / (3.85E2 * 8.96E3) # The thermal conductivity, specific heat capacity, and density of Copper in MKS
dt = 0.1 * (mesh.dx**2 + mesh.dy**2) / (4*k_over_c_rho)
T0 = 273.15 # 0 degree Celsius in Kelvin
T = fipy.CellVariable(mesh, name='T', value=T0+25)
mask_e = mesh.exteriorFaces
T.constrain(T0+25., mask_e)
mask_i = get_mask_of_rect(mesh, 0.25, 0.25, 0.5, 0.5)
T.constrain(T0+30, mask_i)
eq = fipy.TransientTerm() == fipy.DiffusionTerm(coeff=k_over_c_rho)
viewer = fipy.MatplotlibViewer(vars=[T], datamin=0, datamax=400)
plt.ioff()
viewer._plot()
plt.plot(*mesh.faceCenters[:, mask_e], '.g')
plt.plot(*mesh.faceCenters[:, mask_i], '.b')
def update():
for _ in range(10):
eq.solve(var=T, dt=dt)
viewer._plot()
plt.draw()
timer = plt.gcf().canvas.new_timer(interval=50)
timer.add_callback(update)
timer.start()
plt.show()

.constrain() does not work for internal faces (see the warning at the end of Applying internal “boundary” conditions).
You can achieve an internal fixed value condition using sources, however. As a first cut, try
mask_i = get_mask_of_rect(mesh, 0.25, 0.25, 0.5, 0.5)
mask_i_cell = fipy.CellVariable(mesh, value=False)
mask_i_cell[mesh.faceCellIDs[..., mask_i]] = True
largeValue = 1e6
eq = (fipy.TransientTerm() == fipy.DiffusionTerm(coeff=k_over_c_rho)
- fipy.ImplicitSourceTerm(mask_i_cell * largeValue)
+ mask_i_cell * largeValue * (T0 + 30))
This constrains the cells on either side of the faces identified by mask_i to be at T0+30.

I am posting my own answer for future readers.
For boundary conditions for internal faces, you should use the implicit and explicit source terms on the equation, as in the jeguyer's answer.
By using source terms, you don't need to calculate a mask for faces, like this.(The get_mask_of_rect() in my question isn't required.)
T = fipy.CellVariable(mesh, name = 'T', value = T0 + 25)
mask_e = mesh.exteriorFaces
T.constrain(T0 + 25., mask_e)
mask_i_cell = (
(0.25 < mesh.x) & (mesh.x < 0.25 + 0.5) &
(0.25 < mesh.y) & (mesh.y < 0.25 + 0.5)
)
large_value = 1E6
eq = fipy.TransientTerm() == (
fipy.DiffusionTerm(coeff = k_over_c_rho) -
fipy.ImplicitSourceTerm(mask_i_cell * large_value) +
mask_i_cell * (large_value * (T0 + 30) # explicit source
))
viewer = fipy.MatplotlibViewer(vars = [T], datamin = T0, datamax = T0+50)

python - FEniCS How do I convert a tensor expression into a scalar expression?

I'm new to FEniCS and optimization. Trying to run my code I receive the Error
UFLException: Can't add expressions with different shapes.
The Code:
from dolfin import *
from dolfin_adjoint import *
mesh = Mesh("beam.xml") # beam x[0] 0 - 5, x[1] 0 - 1, x[2] 0 - 1
V = VectorFunctionSpace(mesh, "CG", 1)
y = Function(V, name="State") # Trialfunction State
d = y.geometric_dimension() # space dimension
u = Function(V, name="Control") # Trialfunction Control
v = TestFunction(V) # Testfunction
x = SpatialCoordinate(mesh)
alpha = Constant(10e-6)
beta = Constant(0.2)
rho = Constant(7800) # Density of steel [kg/m³]
E = Constant(210000) # Young's modulus steel [MN/m²]
nu = Constant(0.3) # Poisson's ratio steel
lmbda = E*nu/(1+nu)/(1-2*nu) # Lame's constant
mu = E/(2*(1+nu)) # Shear modulus
G = Constant(pi**2) #gravity
F_z = Constant(rho*G) #weight force
F = Constant((0,0,-F_z))
g_z = Constant(2.9575e5)
g = Constant((0,0,g_z))
y_Desired = Constant((0,0,0.0))
top = AutoSubDomain(lambda x: near(x[1], 1))
boundaries = MeshFunction("size_t", mesh, mesh.topology().dim()-1)
boundaries.set_all(0)
top.mark(boundaries, 1)
ds_N = ds(subdomain_data=boundaries) # use in ds_N(1) as in inner(g,v)*ds_N(1)
bottom = AutoSubDomain(lambda x: near(x[1], 0))
boundaries = MeshFunction("size_t", mesh, mesh.topology().dim()-1)
boundaries.set_all(0)
bottom.mark(boundaries, 1)
ds_0 = ds(subdomain_data=boundaries) # use in ds_0(1) as in beta(abs(u))*ds_O(1)
def bottom(x, on_boundary):
return(on_boundary and near (x[1], 0.0))
bc_bottom = DirichletBC(V, u, bottom)
def left(x, on_boundary):
return(on_boundary and near (x[0], 0.0))
bc_left = DirichletBC(V, Constant((0, 0, 0)), left)
def right(x, on_boundary):
return(on_boundary and near (x[0], 5.0))
bc_right = DirichletBC(V, Constant((0, 0, 0)), right)
bcs = [bc_left, bc_right, bc_bottom]
def epsilon(y):
return 0.5*(nabla_grad(y) + nabla_grad(y).T)
def sigma (y):
return lmbda*div(y)*Identity(d) + 2*mu*epsilon(y)
S = inner(sigma(y),epsilon(v))*dx - inner((F+u),v)*dx - inner(g,v)*ds_N(1)
solve(S == 0, y, bcs)
J = assemble((0.5*inner(y- y_Desired, y - y_Desired))*dx + alpha / 2 * inner((F+u), v)* dx + beta*abs(u)*ds_0)
I tried to make y_Desired also 3D but then I receive UFLException: Can only integrate scalar expressions. The integrand is a tensor expression with value shape (3,) and free indices with labels ().
Can anyone help me on this?

How to find minimum value from an array

I'm new to python so the code may not be the best. I'm trying to find the minimum Total Cost (TotalC) and the corresponding m,k and xM values that go with this minimum cost. I'm not sure how to do this. I have tried using min(TotalC) however this gives an error within the loop or outside the loop only returns the value of TotalC and not the corresponding m, k, and xM values. Any help would be appreciated. This section is at the end of the code, I have included my entire code.
import numpy as np
import matplotlib.pyplot as plt
def Load(x):
Fpeak = (1000 + (9*(x**2) - (183*x))) *1000 #Fpeak in N
td = (20 - ((0.12)*(x**2)) + (4.2*(x))) / 1000 #td in s
return Fpeak, td
#####################################################################################################
####################### Part 2 ########################
def displacement(m,k,x,dt): #Displacement function
Fpeak, td = Load(x) #Load Function from step 1
w = np.sqrt(k/m) # Natural circular frequency
T = 2 * np.pi /w #Natural period of blast (s)
time = np.arange(0,2*T,0.001) #Time array with range (0 - 2*T) with steps of 2*T/100
zt = [] #Create a lsit to store displacement values
for t in time:
if (t <= td):
zt.append((Fpeak/k) * (1 - np.cos(w*t)) + (Fpeak/(k*td)) * ((np.sin(w*t)/w) - t))
else:
zt.append((Fpeak/(k*w*td)) * (np.sin(w*t) - np.sin(w*(t-td))) - ((Fpeak/k) * np.cos(w*t)))
zmax=max(zt) #Find the max displacement from the list of zt values
return zmax #Return max displacement
k = 1E6
m = 200
dt = 0.0001
x = 0
z = displacement(m,k,x,dt)
###################################################################################
############### Part 3 #######################
# k = 1E6 , m = 200kg , Deflection = 0.1m
k_values = np.arange(1E6, 7E6, ((7E6-1E6)/10)) #List of k values between min and max (1E6 and 7E6).
m_values = np.arange(200,1200,((1200-200)/10)) #List of m values between min and max 200kg and 1200kg
xM = []
for k in k_values: # values of k
for m in m_values: # values of m within k for loop
def bisector(m,k,dpoint,dt): #dpoint = decimal point accuracy
xL = 0
xR = 10
xM = (xL + xR)/2
zmax = 99
while round(zmax, dpoint) !=0.1:
zmax = displacement(m,k,xM,dt)
if zmax > 0.1:
xL = xM
xM = (xL + xR)/2
else:
xR = xM
xM = (xL + xR)/2
return xM
xM = bisector(m, k, 4, 0.001)
print('xM value =',xM)
####################################################################################
def cost (m,k,xM):
Ck = np.array(900 + 825*((k/1E6)**2) - (1725*(k/1E6)))
Cm = np.array(10*m - 2000)
Cx = np.array(2400*((xM**2)/4))
TotalC = Ck + Cm + Cx
print(TotalC)
print(min(TotalC))
return TotalC
TotalC = cost(m, k, xM)
print([xM, m, k, TotalC])

You want this:
minIndex = TotalC.argmin()
Now you use that numeric index into all your arrays TotalC, Ck, Cm and Cx.

find the index of the min value in total and the index should be same for corresponding arrays.

Taylor Series for sine using functions

I've created a fatorial function that would help me calculate the taylor expansion of sine. There are no evident mistakes in my code, but the returned value is wrong. That's my code:
PI = 3.14159265358979323846
def fatorial(n):
fatorial = 1
for i in range(1,n+1,1):
fatorial = fatorial * i
return fatorial
def seno(theta):
n = 1
k = 3
eps = 10**-10
theta = (theta*PI)/180
x = ((-1)**n)*((theta)**k)
y = fatorial(k)
while x/y > eps or x/y < -eps:
theta = theta + (x/y)
n = n + 1
k = k + 2
x = ((-1)**n) * ((theta)**k)
y = fatorial(k)
return theta

You are summing theta in the while-loop and therefore using an adjusted angle for the next elements of the Taylor series. Just add a thetas (for the sum) like so (I have taken the liberty to add some performance improvements, no need to recalculate the full factorial, also calculated first elements explicitly and changed the limit check to avoid recalculations of x/y):
import math
PI = 3.14159265358979323846
def fatorial(n):
fatorial = 1
for i in range(1,n+1,1):
fatorial = fatorial * i
return fatorial
def seno(theta):
n = 1
k = 3
#eps = 10**-10
theta = (theta*PI)/180
thetas = theta # <- added this
x = -theta*theta*theta
y = 6
#while x/y > eps or x/y < -eps:
while k < 14:
thetas = thetas + (x/y) # sum thetas
n = n + 1
k = k + 2
#x = ((-1)**n) * ((theta)**k)
x = ((-1)**(n%2)) * ((theta)**k) # but use the original value for the series
#y = fatorial(k)
y *= k * (k-1)
return thetas # return the sum
if __name__ == '__main__':
print(seno(80), math.sin(8*PI/18))
this results in
0.984807753125684 0.984807753012208

Python generate all combinations of directions including diagonals in 3 dimensions

I want to generate all directions from a point in a 3D grid, but I can't quite get my head around the next bit. For the record it's all stored in a single list, so I need some maths to calculate where the next point will be.
I only really need 3 calculations to calculate any of the 26 or so different directions (up, up left, up left forwards, up left backwards, up right, up right forwards, etc), so I decided to work with X, Y, Z, then split them into up/down left/right etc, to then get the correct number to add or subtract. Generating this list to get the maths working however, seems to be the hard bit.
direction_combinations = 'X Y Z XY XZ YZ XYZ'.split()
direction_group = {}
direction_group['X'] = 'LR'
direction_group['Y'] = 'UD'
direction_group['Z'] = 'FB'
So basically, using the below code, this is the kind of stuff I'd like it to do, but obviously not have it hard coded. I could do it in a hacky way, but I imagine there's something really simple I'm missing here.
#Earlier part of the code to get this bit working
#I've also calculated the edges but it's not needed until after I've got this bit working
grid_size = 4
direction_maths = {}
direction_maths['U'] = pow(grid_size, 2)
direction_maths['R'] = 1
direction_maths['F'] = grid_size
direction_maths['D'] = -direction_maths['U']
direction_maths['L'] = -direction_maths['R']
direction_maths['B'] = -direction_maths['F']
#Bit to get working
starting_point = 25
current_direction = 'Y'
possible_directions = [direction_group[i] for i in list(current_direction)]
for y in list(possible_directions[0]):
print starting_point + direction_maths[y]
# 41 and 9 are adjacent on the Y axis
current_direction = 'XYZ'
possible_directions = [direction_group[i] for i in list(current_direction)]
for x in list(possible_directions[0]):
for y in list(possible_directions[1]):
for z in list(possible_directions[2]):
print starting_point + direction_maths[x] + direction_maths[y] + direction_maths[z]
# 44, 36, 12, 4, 46, 38, 14 and 6 are all adjacent on the corner diagonals
Here's a general idea of how the grid looks with the list indexes (using 4x4x4 as an example):
________________
/ 0 / 1 / 2 / 3 /
/___/___/___/___/
/ 4 / 5 / 6 / 7 /
/___/___/___/___/
/ 8 / 9 /10 /11 /
/___/___/___/___/
/12 /13 /14 /15 /
/___/___/___/___/
________________
/16 /17 /18 /19 /
/___/___/___/___/
/20 /21 /22 /23 /
/___/___/___/___/
/24 /25 /26 /27 /
/___/___/___/___/
/28 /29 /30 /31 /
/___/___/___/___/
________________
/32 /33 /34 /35 /
/___/___/___/___/
/36 /37 /38 /39 /
/___/___/___/___/
/40 /41 /42 /43 /
/___/___/___/___/
/44 /45 /46 /47 /
/___/___/___/___/
________________
/48 /49 /50 /51 /
/___/___/___/___/
/52 /53 /54 /55 /
/___/___/___/___/
/56 /57 /58 /59 /
/___/___/___/___/
/60 /61 /62 /63 /
/___/___/___/___/
Edit: Using the answers mixed with what I posted originally (wanted to avoid converting to and from 3D points if possible), this is what I ended up with to count the number of complete rows :)
def build_directions():
direction_group = {}
direction_group['X'] = 'LR'
direction_group['Y'] = 'UD'
direction_group['Z'] = 'FB'
direction_group[' '] = ' '
#Come up with all possible directions
all_directions = set()
for x in [' ', 'X']:
for y in [' ', 'Y']:
for z in [' ', 'Z']:
x_directions = list(direction_group[x])
y_directions = list(direction_group[y])
z_directions = list(direction_group[z])
for i in x_directions:
for j in y_directions:
for k in z_directions:
all_directions.add((i+j+k).replace(' ', ''))
#Narrow list down to remove any opposite directions
some_directions = all_directions
opposite_direction = all_directions.copy()
for i in all_directions:
if i in opposite_direction:
new_direction = ''
for j in list(i):
for k in direction_group.values():
if j in k:
new_direction += k.replace(j, '')
opposite_direction.remove(new_direction)
return opposite_direction
class CheckGrid(object):
def __init__(self, grid_data):
self.grid_data = grid_data
self.grid_size = calculate_grid_size(self.grid_data)
self.grid_size_squared = pow(grid_size, 2)
self.grid_size_cubed = len(grid_data)
self.direction_edges = {}
self.direction_edges['U'] = range(self.grid_size_squared)
self.direction_edges['D'] = range(self.grid_size_squared*(self.grid_size-1), self.grid_size_squared*self.grid_size)
self.direction_edges['R'] = [i*self.grid_size+self.grid_size-1 for i in range(self.grid_size_squared)]
self.direction_edges['L'] = [i*self.grid_size for i in range(self.grid_size_squared)]
self.direction_edges['F'] = [i*self.grid_size_squared+j+self.grid_size_squared-self.grid_size for i in range(self.grid_size) for j in range(self.grid_size)]
self.direction_edges['B'] = [i*self.grid_size_squared+j for i in range(self.grid_size) for j in range(self.grid_size)]
self.direction_edges[' '] = []
self.direction_maths = {}
self.direction_maths['D'] = pow(self.grid_size, 2)
self.direction_maths['R'] = 1
self.direction_maths['F'] = self.grid_size
self.direction_maths['U'] = -self.direction_maths['D']
self.direction_maths['L'] = -self.direction_maths['R']
self.direction_maths['B'] = -self.direction_maths['F']
self.direction_maths[' '] = 0
def points(self):
total_points = defaultdict(int)
opposite_directions = build_directions()
all_matches = set()
#Loop through each point
for starting_point in range(len(self.grid_data)):
current_player = self.grid_data[starting_point]
if current_player:
for i in opposite_directions:
#Get a list of directions and calculate movement amount
possible_directions = [list(i)]
possible_directions += [[j.replace(i, '') for i in possible_directions[0] for j in direction_group.values() if i in j]]
direction_movement = sum(self.direction_maths[j] for j in possible_directions[0])
#Build list of invalid directions
invalid_directions = [[self.direction_edges[j] for j in possible_directions[k]] for k in (0, 1)]
invalid_directions = [[item for sublist in j for item in sublist] for j in invalid_directions]
num_matches = 1
list_match = [starting_point]
#Use two loops for the opposite directions
for j in (0, 1):
current_point = starting_point
while current_point not in invalid_directions[j]:
current_point += direction_movement*int('-'[:j]+'1')
if self.grid_data[current_point] == current_player:
num_matches += 1
list_match.append(current_point)
else:
break
#Add a point if enough matches
if num_matches == self.grid_size:
list_match = tuple(sorted(list_match))
if list_match not in all_matches:
all_matches.add(list_match)
total_points[current_player] += 1
return total_points

Here's basically the same thing that #AnnoSielder did, but makes use of itertools to reduce the amount of code.
from itertools import product
# Get a list of all 26 possible ways to move from a given coordinate in a 3 coordinate system.
base_deltas = filter(lambda point: not all(axis ==0 for axis in point), list(product([-1, 0, 1], repeat=3)))
# Define your max axis length or your grid size
grid_size = 4
# Simple function that applys the deltas to the given coordinate and returns you the list.
def apply_deltas(deltas, coordinate):
return [
(coordinate[0]+x, coordinate[1]+y, coordinate[2]+z)
for x, y, z in deltas
]
# This will determine whether the point is out of bounds for the given grid
is_out_of_bounds = lambda point: all(0 <= axis < grid_size for axis in point)
# Define your point, in this case it's block #27 in your example
coordinate = [3, 2, 1]
# Apply the deltas, then filter using the is_out_of_bounds lambda
directions = filter(is_out_of_bounds, apply_deltas(base_deltas, coordinate))
# directions is now the list of 17 coordinates that you could move to.

Don't make thinks unnecessary complicated. Do not describe a point in 3 dimensions with 1 number - 3 coordinates means 3 numbers.
Should be something like this:
numb = 37
cube_size = 4
# convert to x - y - z
start = [0, 0, 0]
start[2] = numb / cube_size ** 2
numb = numb % cube_size ** 2
start[1] = numb / cube_size
start[0] = numb % cube_size
for x in [-1, 0, 1]:
current_x = start[0] + x
for y in [-1, 0, 1]:
current_y = start[1] + y
for z in [-1, 0, 1]:
current_z = start[2] + z
#reconvert
convert = current_x + current_y * cube_size + current_z * cube_size ** 2
print("x: " + str(current_x) + " y: " + str(current_y) + " z: " + str(current_z) + " => " + str(convert))
Simply generate your x/y/z-coordinate, then run all possibilities of add -1/0/1 to these coordinates and re-convert to your number in the grid.