With the ever increasing importance of UQ in various disciplines and fields, software solutions and libraries for UQ problems get more and more important. Progress and use of UQ techniques relies on the availability of software features and support. This raises interesting questions for the UQ community such as: What are the current properties of available tools? For which classes of problems have they been developed? What methods or algorithms do they provide? What are challenges for UQ software and which resources are required? What are recent improvements? What are the next steps and the long-term goals of the development?
This minisymposium brings together experts for different software in the context of UQ, ranging from tools that ease up individual tasks of UQ (such as surrogate modelling, UQ workflows, dimensionality reduction, data augmentation) up to whole frameworks for solving UQ problems. The minisymposium will foster discussion and exchange of ideas between developers and (prospective) users.
08:30
Cossan Software: Efficient and user-friendly computational tools for dealing with uncertainty
Edoardo Patelli | Strathclyde University | United Kingdom
Show details
Authors:
Edoardo Patelli | Strathclyde University | United Kingdom
Santhosh Santhosh | University of Liverpool | United Kingdom
Matteo Broggi | University of Hannover | Germany
Michael Beer | University of Hannover | Germany
Cossan software represents a collection of tools for dealing with uncertainty. Cossan software is a collaborative development across different universities including Strathclyde University, University of Hannover, University of Liverpool. It offers the most advanced and recent algorithms for performing risk, reliability and uncertainty analysis of complex engineering systems.
COSSAN-X provides a user friendly front end of the Cossan Software. The available tutorials and wizards are designed to guide users through the different steps of the analysis making this tool ideal for industry and for training. COSSAN-X provides straightforward steps to integrate third-party software such as Finite Element Solvers. Real-time computational needs of stochastic problems are taken care of by a parallel computing feature. COSSAN-X provides flexible options for postprocessing and a provision for user defined probability distributions and functions.
OpenCossan represents the open-source computational engine of Cossan software. Released under the LGPL license and available in GitHub, it represents an ideal environment for academics to access the state-of-the-art algorithms in UQ and the most popular tools in reliability analysis including advanced MC methods such as Line Sampling, Subset Simulation, Sequential MC. Recent developments have focused on the inclusion of Bayesian Networks Modelling, Credal Network and robust machine learning approaches and on tools applying Robust Neural Networks.
09:00
- CANCELED - UQTk, a C++/Python Toolkit for Uncertainty Quantification: Overview and Applications
Bert Debusschere | Sandia National Laboratories | United States
Show details
Authors:
Katherine Johnston | Sandia National Laboratories | United States
Khachik Sargsyan | Sandia National Laboratories | United States
Cosmin Safta | Sandia National Laboratories | United States
Bert Debusschere | Sandia National Laboratories | United States
The UQ Toolkit (UQTk) is a collection of libraries, tools and apps for the quantification of uncertainty in numerical model predictions. As one of the software tools offered by the DOE SciDAC FASTMath Institute, UQTk offers intrusive and non-intrusive methods for forward uncertainty propagation, tools for sensitivity analysis, sparse surrogate construction, low-rank-tensor approximations, Bayesian inference via various flavors of MCMC, model error assessment, as well as several other capabilities. The core libraries are implemented in C++ but a Python interface is available for easy prototyping and incorporation in UQ workflows. The talk will give an overview of UQTk capabilities and illustrate its application to complex scientific workflows.
09:30
PolyChaos.jl – An Open Source Julia Package for Orthogonal Polynomials, Quadrature, and Polynomial Chaos Expansion
Tillmann Mühlpfordt | Karlsruhe Institute of Technology | Germany
Show details
Authors:
Tillmann Mühlpfordt | Karlsruhe Institute of Technology | Germany
Frederik Zahn | Karlsruhe Institute of Technology | Germany
Timm Faulwasser | Karlsruhe Institute of Technology | Germany
Veit Hagenmeyer | Karlsruhe Institute of Technology | Germany
Orthogonal polynomials play a lively role in the field of uncertainty quantification, be it for quadrature rules, or polynomial chaos expansions. Reliable, efficient, and open source software that provides easy access to orthogonal polynomials is key in applications such as uncertainty propagation or optimization under uncertainty. We introduce PolyChaos.jl – a software package for orthogonal polynomials, quadrature rules, and polynomial chaos expansions in the Julia programming language. Julia is a trending language dedicated to scientific and technical computing. It combines the readability of interpreted languages with the speed of compiled languages. PolyChaos.jl comes with several well-known orthogonal polynomials, e.g. Hermite, or Jacobi. Given an absolutely continuous non-negative measure PolyChaos.jl also allows to compute the respective orthogonal polynomials, for instance via the Stieltjes procedure, or the Lanczos procedure. Based on the orthogonal polynomials PolyChaos.jl computes the quadrature rules via Gauss/Lobatto/Radau-quadrature. For intrusive polynomial chaos, PolyChaos.jl further provides the tensorized scalar products of the orthogonal basis functions. Besides providing technical details our presentation focuses on the comprehensible documentation and possible future extensions of PolyChaos.jl.
10:00
UQpy: A Python toolkit and development environment for UQ
Aakash Bangalore Satish | Johns Hopkins University | United States
Show details
Authors:
Michael D. Shields | Johns Hopkins University | United States
Dimitrios Giovanis | Johns Hopkins University | United States
Aakash Bangalore Satish | Johns Hopkins University | United States
UQpy (Uncertainty Quantification with Python) is an open source, general-purpose Python toolkit for modeling, quantifying and propagating uncertainty in the simulation of physical and mathematical systems. UQpy is an advanced tool for performing UQ operations, that offers flexibility to the end-user because of its modular nature that is designed to be centered around a set of core capabilities. However, the object-oriented modular structure of the code makes it readily extensible for the integration of new capabilities. Thus it provides an integrated development environment for advanced UQ modelers to build new methods and capabilities. In UQpy, each module contains a set of classes that have the ability to invoke one-another. These modules are centered around a core module (RunModel) that interfaces with deterministic Python or third-party solvers. This allows UQpy to serve as a driver for performing complete non-intrusive uncertainty studies – including pre-processing operations, submission and execution of computational model evaluations in serial or parallel, monitoring and post-processing of results, and iterative/active learning.