OpenTURNS is the Global Uncertainty Processing Methodology vector modelling software developed by the EDF's R&D Uncertainties Network. It offers a wide-ranging set of methods to quantify, propagate and rank uncertainties.
- for quantification, OpenTURNS provides sample-based statistical processing functions, including parametric and non-parametric adjustment tests. It also enables modelling of density linear combination laws and random variables. Random dependencies are modelled using copula theory. Based on the association of marginals and copula, the multivariate law construction mechanism enables complex manipulations of multivariate probability laws (copula extraction from a multivariate law reconstructed by nodes, copula composition, etc.)
- for propagation, OpenTURNS offers analytical calculation methods wherever possible to determine the variable of interest law based on manipulation of characteristic functions, as well as multiple simulation methods and FORM/SORM methods generalised to the elliptical copulas
- for ranking, OpenTURNS evaluates importance factors and sensitivity adapted to the propagation method selected, including Sobol indices
- lastly, OpenTURNS offers the option to build response surfaces, including polynomial chaos expansions
THREE OPTIONS FOR USE
Developed under Unix/Linux to work on Windows, OpenTURNS can be used in three different ways:
- as a C++ library incorporated into a specialist application
- as a python module
- using a graphic interface based on the EFICAS open source software developed by EDF's R&D
COMPATIBILITY, LICENSING AND DOCUMENTATION
OpenTURNS is compatible with SALOME, the CAD-computational-visualisation workbench developed by a partnership that includes EDF's R&D.
The source code of OpenTURNS is subject to LGPL licensing, and its documentation element is covered by the GNU FDL.
OpenTURNS can propagate uncertainties across a broad diversity of models : these models may be written in (almost) any language and at any level of complexity. As a result, very simple models can be written directly in Python. But OpenTURNS can also be interfaced with business code written in C, C++, FORTRAN or other language.
OpenTURNS offers detailed user documentation:
- the business content is covered by the ReferenceGuide, which describes the overall methodology and each method used
- the text interfaces covered by the Use CasesGuide, which contains around 100 Python scripts covering all the Python operators offered
- the User Manual defines the parameters used for the majority of methods
- the ExampleGuide shows how the methodology is implemented using a series of instructional examples
OpenTURNS is the subject of an ITECH training course (in French): "Uncertainty processing using OpenTURNS".