epicf / ef_python Goto Github PK

https://colab.research.google.com/drive/1sXa2tNonqMNLk7NjkUY6cx9_keGS3rG5#scrollTo=HA_RfvC7o1po

Prototype several fast solvers for Poisson's equation.
First stage: 2d region, rectangular region, boundary conditions of the first type.

Needed:
FFT (e.g. https://epubs.siam.org/doi/abs/10.1137/1019071)
Cyclic reduction (as above)
FFT+domain decomposition ( http://num-meth.srcc.msu.ru/zhurnal/tom_2015/pdf/v16r110.pdf )

FFT and Cyclic reduction are in progress;

Может файл для визуализации в jupyterе внести в основную программу? А то у меня он дорабатывается. И в разных папках он разных версий получается. Или папку отдельную сделать?

• Wait until code is more finished
• Finalize package name selection
  • Is ef is too short?
  • Is ef_python informative and readable?
• Update README
• Check and update setup.py

Migration to numpy from lists and Vec3d classes should accelerate the code.
Besides, pyCUDA and pyOpenCL libraries can transfer numpy arrays to GPU semiautomatically.

Я сделал пример где сравниваются пучки с электродами и внешним полем. Нашел два недостатка:

1. Пучок пролетает сквозь анод. В первом расчете
2. После запуска ячейки с командой выполнения программы. Она все посчитала. Все h5 файлы есть в папке, но ячейка почему то в работе знак [*] не исчезает . Приходится останавливать kernel. И продолжать строить графики уже после остановки.

Finish migration of examples to Jupyter notebooks.
Some examples are ready; update wiki pages accordingly.
In the wiki leave only problem statement and short results summary; move details into notebooks.
For each example provide:

• a link to Jupyter notebook in the repo
• a link to Google Colab version
• a command to run console scripts

Current status:
* Single Particle in Free Space
* Single Particle In Uniform Magnetic Field
* Single Particle In Uniform Electric Field
* Ribbon Beam Contour
* Contour of Ribbon Beam In Uniform Magnetic Field
* Axially Symmetric Beam Contour
* Contour of Axially Symmetric Beam In Uniform Magnetic Field
* Potential well of cylindrical beam in tube
* Conducting Sphere Potential
* Child-Langmuir Law for Planar Diode
* Maxwell-Boltzmann Distribution Test
* Off-Axis Ion Extraction
* Acceleration of Ions by Collisions with Electrons

Хотел по образу и подобию сделать чтение из h5 файла как у тебя. Но методы например из Spatial_mesh исполнялись с ошибкой. Всю ночь бился. Только сегодня подумал проверить как у тебя чтение работает. И не смог запустить

ab@ab-work:~/ef_python/examples/single_particle_in_magnetic_field$ python3 ../../main.py long_simulation_0000000.h5
Traceback (most recent call last):
File "../../main.py", line 73, in
main()
File "../../main.py", line 18, in main
dom, continue_from_h5 = construct_domain( config_or_h5_file )
File "../../main.py", line 36, in construct_domain
extract_filename_prefix_and_suffix_from_h5filename( config_or_h5_file )
File "../../main.py", line 60, in extract_filename_prefix_and_suffix_from_h5filename
if len( match.group ) == 1:
TypeError: object of type 'builtin_function_or_method' has no len()

Jupyter and matplotlib are only needed for visualization, but have many dependencies

[ParticlesBehaviourOnBoundaries]
'absorb'
'rebound'
'periodic'

That seems a better name.

This would also concern EfJupyter default ef-running command (and a couple of examples).

There are several typical tasks for visualization, e.g:

• Plot particles positions in 3d
• Plot 2d projection of particles positions
• Plot trajectories of particles with given id
  etc

Define such functions to quickly preview results of simulation.

After calculation the terminal shows:

Saving 3D trajectory plot to "3d.png"
Saving 2D trajectory projection plots to "2d.png"
Saving kinetic energy comparison plot to "kin_en.png"

But at fact the files were not created in the folder of the example.

It seems it is not used anymore.

Странно, что в визуализации концы конуса задаются двумя точками, а в фигуре - одной четверкой координат. Работать не должно по-моему.

Currently there is some code duplication between InnerRegions and ParticleSources
concerning their geometric shape. It seems natural to define separate classes for geometric primitives and utilize them.

The work has been started in https://github.com/epicf/ef_python/tree/geometric-primitives branch.
For each primitive it is necessary to provide at least two methods: a way to check if a point with given coordinates is inside this primitive (to remove particles after collision) and a way to generate a point inside it (for sources).

Besides, annoying problem is to read primitive definitions from config.
There is no natural support for subsections in INI-files so it is necessary to write geometric primitives as strings and then parse those string in class constructors. It has been implemented (though not tested yet), but probably there is a better solution.

Current status:
It is necessary to test initialization of classes from text string. And it is necessary to replace current geometry-related code in inner regions and sources with this new classes. Then it would be necessary to update and test examples.

SI units seem more convenient for current applications.
Some parts of the code and examples have to be transferred.

This has been fixed before, but for some reason this bug is still present in the master branch.

Пример с пушкой пирса. Я задал начальный внутренний радиус конуса катода 0. Но видимо при работе функции "найти координату в области" получилось что несколько ячеек вблизи нуля не являются частью электрода. Частицы пролетают сквозь конус в обратном направлении.

[TimeGrid]
total_time = 3.5e-10
time_save_step = 3.5e-12
time_step_size = 3.5e-12

[SpatialMesh]
grid_x_size = 1.4
grid_x_step = 0.035
grid_y_size = 1.4
grid_y_step = 0.035
grid_z_size = 0.7
grid_z_step = 0.0175

[ParticleSourceCylinder.emit_single_particle]
cylinder_axis_start_x = 0.7
cylinder_axis_start_y = 0.7
cylinder_axis_start_z = 0.2
cylinder_axis_end_x = 0.7
cylinder_axis_end_y = 0.7
cylinder_axis_end_z = 0.21
cylinder_radius = 0.06
initial_number_of_particles = 100
particles_to_generate_each_step = 100
mean_momentum_x = 0.0
mean_momentum_y = 0.0
mean_momentum_z = 1.225e-15
temperature = 0.0
charge = -1.049e-05
mass = 2.142e-23

[InnerRegionConeAlongZ.Cone_Catode]
cone_axis_x = 0.7
cone_axis_y = 0.7
cone_axis_start_z = 0.2
cone_axis_end_z = 0.35
cone_start_inner_radius = 0.0
cone_start_outer_radius = 0.275
cone_end_inner_radius = 0.587
cone_end_outer_radius = 0.6
potential = -3.3

[InnerRegionConeAlongZ.Cone_Anode]
cone_axis_x = 0.7
cone_axis_y = 0.7
cone_axis_start_z = 0.52
cone_axis_end_z = 0.67
cone_start_inner_radius = 0.0
cone_start_outer_radius = 0.275
cone_end_inner_radius = 0.0
cone_end_outer_radius = 0.6
potential = 0

[OutputFilename]
output_filename_prefix = pierce_
output_filename_suffix = .h5

[BoundaryConditions]
boundary_phi_right = 0.0
boundary_phi_left = 0.0
boundary_phi_bottom = 0.0
boundary_phi_top = 0.0
boundary_phi_near = 0.0
boundary_phi_far = 0.0

[ParticleInteractionModel]
particle_interaction_model = PIC

It is necessary to make sure that examples run and produce reasonable results after code updates.
While it is not obvious how to test results, it should be possible to test at least start, restart and output files automatically.

В файле InnerRegionsManager.py в inite 52 строчка

        elif InnerRegionConeAlongZ.is_cone_region(sec_name):
            new_obj.regions.append(
                InnerRegionTube.init_from_config(conf,
                                                 conf[sec_name],
                                                 sec_name,
                                                 spat_mesh))

В cone трубка создается получается