6. Release Notes

6.1. Major Version 3

Warning

The major version number was increased from 2 to 3 as changes were made to the input and output formats of quoFEM app. This means old examples will not be loaded in this version of the tool.

Version 3.5 (Current)

Release date: December. 2023

Highlights

  1. Bayesian calibration of a hierarchical model

Version 3.4

Release date: October. 2023

Highlights

  1. Multi-fidelity Monte Carlo Simulation

Version 3.3

Release date: March. 2023

Highlights

  1. Multimodel uncertainty propagation

  2. Stochastic kriging without replications

  3. Display of correlation coefficients within the input/output dataset

  4. Switching the display order of the UQ method and UQ engine

Version 3.2

Release date: September. 2022

Highlights

  1. Support for a gradient-free optimization and stochastic Kriging

  2. Fast global sensitivity analysis for very high dimensional output (tested on 2 million QoIs)

  3. New option to discard working directories after each model simulation

  4. Support for PLoM on DesignSafe

  5. Significantly enhanced speed of surrogate validation and prediction

  6. None option for FEM

  7. Improved user interface including error bounds of the surrogate prediction

  8. Major renaming:

    • OpenseesPy → python

    • Parameters estimation → deterministic calibration

    • Inverse problem → Bayesian calibration

Version 3.1

Release date: June. 2022

Highlights

  1. New efficient global sensitivity analysis method for high-dimensional output (GSA-PCA)

  2. “Save RVs” and “Save QoIs” buttons were added to the results tab spreadsheet

  3. “NaN” handling option added to SimCenterUQ engine

  4. Improvements to reliability analysis and global sensitivity analysis user interface

  5. Minor bug fixes in the user interface, surrogate modeling, and sensitivity analysis scripts

Version 3.0

Release date: March. 2022

Highlights

  1. New option for surrogate modeling using Probabilistic Learning on Manifolds (PLoM)

  2. Restructured surrogate model scripts

  3. Improvements to the user interface for RV, QoI and RES tabs

  4. Improvements to the message area

  5. Major restructuring of the backend

  6. Minor bug fixes in the user interface, surrogate modeling and sensitivity analysis scripts

  7. Updated example files

6.2. Major Version 2

Version 2.4.1

Release date: Dec. 2021

Highlights

  1. Added ‘file_save’ keyword in dakota.in to not delete paramsDakota.in files

  2. SimCenterUQ RV tab - preventing path strings from being deleted when “choose” is clicked (dataset inputs)

  3. SimCenterUQ checks if Python packages are missing in the environment and shows an error message if needed

  4. Minor fixes in surrogate UI (nugget values option should not show up by default, RVs should be uniform by default)

  5. A fix to prevent the mixed use of slash/backslash when printing a path

  6. Parameter values are passed to the log-likelihood script when using the UCSD_UQ engine

Version 2.4.0

Release date: Oct. 2021

Highlights

  1. New forward propagation method in SimCenterUQ to import existing sample sets (e.g. samples obtained by MCMC)

  2. New multi-fidelity surrogate modeling option in SimCenterUQ

  3. Local/remote parallel computing support for SimCenterUQ methods

  4. Visualization improved for surrogate results

  5. More adaptive design of experiment options added for surrogate modeling

  6. Nugget optimization options added for surrogate modeling

  7. Minor improvements and bug fixes

Version 2.3

Release date: May 2021

Highlights

  1. Data for calibration methods (DREAM, TMCMC, parameter estimation) required to be provided in a file

  2. Option to supply a covariance structure for error in Bayesian calibration methods

  3. Option to calibrate values of multipliers on error covariance structure in Bayesian calibration methods

  4. Log-likelihood function specification made optional for TMCMC

Version 2.2

Release date: Oct. 2020

Highlights

  1. Included new sensitivity method: probability model-based global sensitivity analysis (PM-GSA)

  2. Included new Bayesian calibration method: transitional Markov chain Monte Carlo (TMCMC)

  3. Option to allow users to include their own UQ engine

  4. Option to allow users to include their own FEM engine

  5. Changes to UI to reduce wasted space

Version 2.0

Release date: Sept. 2019

Highlights

  1. Forward uncertainty: Importance Sampling, Gaussian Process Regression

  2. Reliability: FORM and SORM

  3. Sensitivity with Monte Carlo or LHS

  4. Parameter Estimation