Skip to content

v3.1.2

Latest
Compare
Choose a tag to compare
@friedenhe friedenhe released this 18 Jun 18:11
· 11 commits to main since this release
3492f59

Changes

  • Added primal and adjoint solvers for DARhoPimpleFoam 217f5e1. This solver is still in the beta state.
  • Changed the intermediate optimization folder format to 0.0001, 0.0002, etc. 5f2cc15
  • Allowed multi betaFI by using new names such as betaFINuTilda. 7570ed7
  • Added an option to use step-averaged states for steady-state solvers. 1b56851 This is for steady-state solvers with oscillating flow states due to separation.
  • Enabled using objective function's standard deviation as convergence criterion for steady-state solvers. d58cd15
  • Added the radial basis function for regressionModel. Also added the ReLU activation function. 35e0a4c
  • Allowed parallel field inversion cases with v3 run scripts. d0379b0
  • Allowed saving flow features to the disk for field inversion. 8afd2ab
  • Added TensorFlow callback functions for the regModel 304957a
  • Enabled using multiple regression models for computing multiple augmented fields in field inversion. be232cb
  • Allowed prescribing heat source for DAHeatTransferFoam. 8a9c93a
  • Added an option to use the latest flow fields instead of reading initial fields from the 0 folders for unsteady solvers. eb6d1b3

Dependency update

  • Updated the pyOFM version to 1.2.2 with new APIs for read/write fields.

Backward-incompatible changes

  • The naming convention for optimization intermediate folders has changed to 0.0001, 0.0002, etc. This will not change how we run the case, but it may create backward compatibility for some post-processing scripts that need to read the intermediate folders.
  • The augmented flow variable's name has changed from betaFI to more specific variable names, such as betaFINuTilda, betaFIOmega, etc. Users need to change betaFI in their runScript.py to the new naming convention.
  • The regressionModel dict has a new interface that supports multiple regression models. See an example from here
  • The tensorflow dict has a new interface that supports multiple regression model. See here