Towards the second year milestone: the UQLab development roadmap

Abstract

Uncertainty quantification is rapidly becoming a well-established topic in many fields of engineering and applied sciences. However, the lack of proper software tools that combine generality, accessibility, ease of use and state-of-the-art techniques has somewhat hindered the diffusion of many powerful tools to a wider scientific audience. Under these premises, the UQLab software project started in ETH Zurich in early 2013 with the aim of providing a powerful, modular, robust and user-friendly platform for uncertainty quantification across disciplines. Following the 3-step general framework for uncertainty quantification described in Sudret (2007), UQLab is designed to help researchers to solve a large class of uncertainty quantification problems by providing a natural environment to pose them, as well as advanced algorithms for their solution (Marelli and Sudret, 2014). The first beta version of UQLab was released in July 2015. Among its features, it included the joint probabilistic representation of random variables based on the copula formalism, a module to connect to various kinds of existing computational models (e.g. Matlab functions or external software through wrappers), advanced metamodelling techniques (e.g. sparse polynomial chaos expansions and Gaussian process regression), state-of-the-art sensitivity analysis tools (e.g. Sobol’ indices, Morris elementary effects, etc.) and a Reliability analysis module (e.g. FORM/SORM, Importance Sampling, adaptive surrogate-based AK-MCS, etc.). Close to the two years milestone and after the open source release of the code of its scientific modules, a number of open topics and development perspectives are presented in this contribution. Among them: the extension of the available metamodelling tool to comprise low rank canonical representations, PC-Kriging, support vector regression and classification, as well as the extension of the probabilistic input representation with random fields, high dimensional copula inference, etc. The planned launch of a global UQLab community to further foster the interdisciplinary character of its development will be also discussed. References [1] Sudret, B. (2007). Uncertainty propagation and sensitivity analysis in mechanical models - Contributions to structural reliability and stochastic spectral methods. Habilitation à diriger des recherches, Université Blaise Pascal, Clermont-Ferrand, France. [2] Marelli, S. & Sudret (2014), B. UQLab: A framework for uncertainty quantification in Matlab. Proc. 2nd Int. Conf. on Vulnerability, Risk Analysis and Management (ICVRAM2014), Liverpool, United Kingdom, 2554-2563.