* Fixed bug in float division * Exception check and demo added (printhead unsteady control) * Bug fix in optimization bnds * Only half of the full energy is considered, missed `/ 2.0` for free surface energy * Minor changes and Taylor test * Fixed bounds creation step * Free surface model update: consistent with CN-time stepping, fixed noisy behaviour at kappa=0 * Unsteady adjoint solver now should be CN-consistent * Taylor testing of unsteady control w/ and without free surface * Updated unsteady solvers for printhead (and comments removed in tube control) * see prev commit message * Taylor tests for pure acoustics * Starting optimization from coarse (width=2.0) solution * Finer space nearly optimal solution * Added reduced_basis option * better value for control * Now mass matrices are scaled by the step size (dt) * Resets the optimization, window=default (2.0) * Bug fix with step size multiplier * Small docstring and data additions * tube control taylor testing * Bug fix in parse_real_expression * Changed window * ignore .dat files * BUG: basis generation fix * tmp test for energy visualization * DEV: now 1D domains are supported in weak forms * DEV: 1D domain mesher * DEMO: 1D domain unsteady adjoint optimizer * gitignore * RUN: reduced basis run * RUN: one direct run * RUN: single run tests * DEV: kinetic energy flux through boundary calculation * MAINT: removing unused data during unsteady run. the last point must still be present * DEV: optimization with small piezo boundary L = 0.4 * REV: revert the remove unused data in TimeSeries commit * RUN: visualization of short actuator, controlled case * MAINT: removed TimeSeries storage of full state during direct and adjoint runs (Issue #1) * DEV: custom folder name for solver visualization * DEV: reading data from json. Customized output/run mode * MAINT: small naming corrections * FIX: fixed waveform window nondimensional size * MAINT: using csvwriter instead of plain write * MAINT: csvwriter - removed redundant str() statements * MAINT: now only log energy every step if run mode is single run * MAINT: removed unused code * DEV: added argparse to pass input config file * MAINT: removed (hid) the assert statements * FIX: numpy to list csv output bug * MAINT: added missing kwarg in adjoint run: plot_every * MAINT: added lower/upper bounds for control space * MAINT: refactored optimization method in mixin to kwarg * DEV: added algorithm parser * DEV: symmetric printhead factory class * MAINT: moved `Constants` to free surface model module * MAINT: using printhead geometry assembler instead of explicit instantiation * MAINT: moved argparser to new module * MAINT: refactored the assembler common methods to base class * DEV: added symmetric injector assembler * DEV: added geometry registry * MAINT: taylor test fix * FIX: taylor test magnitude * FIX: reduce taylor test step * taylor test * taylor test initial control magnitude * REG: added small slip wall next to the actuator * taylor test setup: step size * DEV: Base geometry assembler uses keyword "metadata" for boundary conditions * MAINT: symmetric injector class refactored to allow zero slip region near the actuator * MAINT: only apply force boundary condition if it's normal force boundary * DEV: more explicit formulation of injector domain * MAINT: small corrections in injector mesh data * DATA: example injector domain * DEV: added final time data to experiment_id * DEV: install firecrest as a CLI program * MAINT: explicit kwarg for maximum number of iterations in `minimize` * MAINT: time series updates (no need for small grid) * DEV: major refactor * FIX: fixed midpoint definition * DEV: added logger factory and basic logger * MAINT: error messages improved * DEV: added kwarg for optimization method to be passed at runtime * DEV: unsteady optimizer - first attempt to make it reusable * DEV: added free surface model factory * DEV: finished first version of application class, with limited capabilities of running only unsteady solver * DEBUG: caching error? * BUGFIX: mutated initial state? * BUGFIX: mutated initial state? * MAINT: renamed `initial_state` to avoid property interference * MAINT: removed print * MAINT: import just pi from math; creating copy of dict for further processing * MAINT: removed commented lines * DEV: Inflow model * application update * API update to time storage * acoustic weakform: added mass flow and stress * solver update * old code * logging dissipation * dissipation form * Single run now also performs adjoint computation and logs the sensitivity * MAINT: minor additions to the eigenvalue example * FIX: using (P1, P2, P2) fspaces instead of linear for Temperature and quadratic for pressure * Printhead demo with nonlinear spectral BC * Printhead spectrum + volumetric dissipation * Fix: Pe -> Pr * DEV: added `inhom_nozzle_impedance` boundary condition * MAINT: small refactor (STY) * DEMO: frequency dependency * MAINT: small improvements * Dev/one dim forms (#5) * MAINT: api updates * taylor testing * taylor tests * latest * STY: split lines in comment code * MAINT: increased maxiter number to 200 * DEV: better taylor test tool * 1d waveform visualization * trying small timesteps to avoid oscillations * FIX: don't plot! * MANT: removed tmp files * MAINT: removed tmp files * removed tmp files * unit domain waveform optimiztation * firecrest.api + improve demos (#7) * MAINT: DEMO: fixed Pe->Pr * DEV: explicitly use "mumps" in `SpectralTVAcousticSolver` * MAINT: spectral acoustic demo updated * DEMO: improved domain demo * MAINT: changed `el_size` in geometry demo * DEMO: eigenproblem in triangle * DEV: control nof converged modes in solver.solve(number_of_modes) * DEV: added api file * MAINT: api improvements * MAINT: use API imports * Fix mesh properties (#8) * MAINT: DEMO: fixed Pe->Pr * DEV: explicitly use "mumps" in `SpectralTVAcousticSolver` * MAINT: spectral acoustic demo updated * DEMO: improved domain demo * MAINT: changed `el_size` in geometry demo * DEMO: eigenproblem in triangle * DEV: control nof converged modes in solver.solve(number_of_modes) * DEV: added api file * MAINT: api improvements * MAINT: use API imports * FIX: remove redundant properties in the Geometry class * Helper function for constructing spatial and temporal forms (#9) * MAINT: refactor the pointwise forms into separate private functions * MAINT: remove demos that are not yet ready / too outdated (#11) * Parabolic inflow + refactoring (#12) * DEV: refactoring + added Parabolic Inflow Model * MAINT: explicit arguments in the `SurfaceModelFactory` * DEV: more generic interface in the `InflowModelFactory` * DEV: use updated signatures in the app / optimizer * MAINT: remove commented code * minor fixes * DEV: added `weighted_stress` method to the acoustic weak forms * DEV: use weighted stress to evaluate adjoint sensitivity * MAINT: refactor the `UnsteadyOptimizationSolver` run signature * DEV: change the `InflowModelFactory.create_model` to use type string as input argument * STY: minor change to reduce overall diff * MAINT: remove commented old code * FIX: use correct (2nd order) expression for surface energy (#13) * Refinement option for SimpleDomain (#14) * DEV: do not add unnecessary boundary points in 1st order (linear) boundary elements * DEV: `SimpleDomain` now can be refined using the "refinement level" argument * MAINT: use logging instead of print * FIX: use correct gamma value for liquids (water) (#15) * LOG: log number of degrees of freedom (#16) * Minor improvements (#17) * DEV: add a `direct_adjoint` to run direct-adjoint simulation, and `single_run` for direct only * MAINT: FirecrestApp to return from `run` * Refactor mesh generation (#18) * Change project requirements * DEV: refactor mesh generation, get rid of `gmsh` exe dependency * update README * update readme (#19) * WARN FIX: downgrade h5py to 2.10 (#21)

Memleak in unsteady run #1

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions