Open
Description
Contact Details
No response
What happened?
When trying to read out the energy from the obstacle example like this
import torch
import lettuce as lt
import matplotlib.pyplot as plt
import numpy as np
"""
Context definitions.
The context defines the default device (cpu or cuda) and datatype (e.g.,
float32 for single, float64 for double precision).
Native CUDA is currently not supported for the anti-bounce-back outlet.
"""
context = lt.Context(torch.device("cuda:0"), use_native=False)
"""
Flow definitions.
We need
1. the resolution in x and y direction
2. the Reynolds number (i.e., how fast the flow behaves compared to the
object's length and fluid's viscosity)
3. the Mach number (i.e., how fast the flow is compared to speed of sound;
Ma=0.3 is stable, above is discouraged)
4. the physical domain length in x-direction (this defines how lattice units
scale to physical units)
to initialize the Obstacle flow object.
"""
nx = 200
ny = 100
Re = 300
Ma = 0.01
lx = 1
flow = lt.Obstacle(context, [nx, ny], reynolds_number=Re, mach_number=Ma,
domain_length_x=lx)
"""
Per default, lt.Obstacle has no solid. It is stored in flow.mask as a fully
False numpy array.
To add a solid, we set some mask values to True by getting the domain extends
from flow.grid and creating a boolean array from it.
For a circle, just use a boolean function. Otherwise, you may as well use the
array indices.
"""
x, y = flow.grid
r = .05*y.max() # radius
x_c = 0.3*x.max() # center along x
y_c = 0.5*y.max() # center along y
flow.mask = ((x - x_c) ** 2 + (y - y_c) ** 2) < (r ** 2)
"""
To show 2D images, you need to rotate the outputs. This is because in lettuce,
the first axis is downstream, while for imshow it is vertical.
"""
plt.imshow(context.convert_to_ndarray(flow.mask.t()))
plt.show()
"""
Collision definition.
The collision is usually BGK (low dissipation, but may be unstable) or KBC
(higher dissipation, but generally stable). BGK is preferred for converging
flows, KBC is preferred for driven flows in smaller domains (where energy
conversation plays a smaller role, but gradients may be higher).
"""
collision = lt.KBCCollision2D(tau=flow.units.relaxation_parameter_lu)
"""
Simulation object setup.
"""
simulation = lt.Simulation(flow=flow, collision=collision, reporter=[])
"""
Reporters.
- Reporter objects are used to extract information later on or during the
simulation.
- They can be created as separate objects when required later
(see 01_example4convergence.py).
"""
energyreporter = lt.ObservableReporter(lt.IncompressibleKineticEnergy(flow),
interval=50)
simulation.reporter.append(energyreporter)
"""
Run num_steps iterations of the simulation.
This can be done repeatedly (see 02_converging_flows.py).
"""
mlups = simulation(num_steps=4000)
print("Performance in MLUPS:", mlups)
"""
Before or after simulation call, physical values can be extracted from the
flow.
Alternatively, the reporters can be drawn from the simulation.reporters list
(see 01_example4convergence.py)
"""
print("simulation.reporter[0].out ",simulation.reporter[0].out)
energy = np.array(simulation.reporter[0].out)
print(energy.shape)
plt.plot(energy[:,1],energy[:,2])
plt.title('Kinetic energy')
plt.xlabel('Time')
plt.ylabel('Energy in physical units')
plt.show()
u = context.convert_to_ndarray(flow.u_pu)
u_norm = np.linalg.norm(u, axis=0).transpose()
plt.imshow(u_norm)
plt.title('Velocity after simulation')
plt.colorbar()
plt.tight_layout()
plt.show()
the following error occurs
[103](vscode-notebook-cell:?execution_count=15&line=103) energy = np.array(simulation.reporter[0].out)
[104](vscode-notebook-cell:?execution_count=15&line=104) print(energy.shape)
--> [105](vscode-notebook-cell:?execution_count=15&line=105) plt.plot(energy[:,1],energy[:,2])
[106](vscode-notebook-cell:?execution_count=15&line=106) plt.title('Kinetic energy')
[107](vscode-notebook-cell:?execution_count=15&line=107) plt.xlabel('Time')
IndexError: too many indices for array: array is 0-dimensional, but 2 were indexed
The reason for that is that simulation.reporter[0].out has type ``<ipykernel.iostream.OutStream at 0x7fee7ea2e200>``` and canot be read in as a numpy array.
Version
0.2.0
What operating system are you seeing the problem on?
No response
Activity