Nowadays computational models are standard tool of analysis in many different areas of applied sciences (e.g. physics, engineering, biology, etc) as they allow to explore, with great flexibility and at relatively low cost when compared to an approach that requires an experimental part, the behavior of a (possibly) complex system of interest. However, any computational model is uncertain with respect to the system of interest, due to variabilities on its parameters, surrounding conditions, etc (data uncertainties) and, mainly, because of the assumptions made on its conception that may not be in agreement with reality (model uncertainties). Take into account these uncertainties and quantify their impact on the system response is often a challenging task, that is essential for: (i) specify a reliability region around numerical simulation results; (ii) certify risk decisions; (iii) design components with low sensitivity to parameter variation (robust design); (iv) validate physical models; etc. This presentation aims to expose the types of uncertainties that are inherent to physical systems, how to consistently construct a stochastic model of uncertainties through a probabilistic approach, and to present some examples of applications.
*Texto informado pelo autor.
Americo Cunha is an Assistant Professor of Applied Mathematics at Rio de Janeiro State University (UERJ), co-founder and coordinator of the Nucleus of Modeling and Experimentation with Computers – NUMERICO. He has B. Eng., M.Sc. and D.Sc. degrees in Mechanical Engineering from Pontifical Catholic University of Rio de Janeiro (PUC-Rio) in Brazil, where he also obtained a B.Sc. degree in Applied Mathematics. His research interests include computational science and engineering, uncertainty quantification, nonlinear dynamics, inverse problems and industrial mathematics.