• search hit 7 of 31
Back to Result List

## ESTIMATING UNCERTAINTIES FROM INACCURATE MEASUREMENT DATA USING MAXIMUM ENTROPY DISTRIBUTIONS

• Modern engineering design often considers uncertainties in geometrical and material parameters and in the loading conditions. Based on initial assumptions on the stochastic properties as mean values, standard deviations and the distribution functions of these uncertain parameters a probabilistic analysis is carried out. In many application fields probabilities of the exceedance of failure criteriaModern engineering design often considers uncertainties in geometrical and material parameters and in the loading conditions. Based on initial assumptions on the stochastic properties as mean values, standard deviations and the distribution functions of these uncertain parameters a probabilistic analysis is carried out. In many application fields probabilities of the exceedance of failure criteria are computed. The out-coming failure probability is strongly dependent on the initial assumptions on the random variable properties. Measurements are always more or less inaccurate data due to varying environmental conditions during the measurement procedure. Furthermore the estimation of stochastic properties from a limited number of realisation also causes uncertainties in these quantities. Thus the assumption of exactly known stochastic properties by neglecting these uncertainties may not lead to very useful probabilistic measures in a design process. In this paper we assume the stochastic properties of a random variable as uncertain quantities caused by so-called epistemic uncertainties. Instead of predefined distribution types we use the maximum entropy distribution which enables the description of a wide range of distribution functions based on the first four stochastic moments. These moments are taken again as random variables to model the epistemic scatter in the stochastic assumptions. The main point of this paper is the discussion on the estimation of these uncertain stochastic properties based on inaccurate measurements. We investigate the bootstrap algorithm for its applicability to quantify the uncertainties in the stochastic properties considering imprecise measurement data. Based on the obtained estimates we apply standard stochastic analysis on a simple example to demonstrate the difference and the necessity of the proposed approach.  • Volltext Document Type: Conference Proceeding Thomas MostORCiDGND https://doi.org/10.25643/bauhaus-universitaet.2873Cite-Link https://nbn-resolving.org/urn:nbn:de:gbv:wim2-20170314-28732Cite-Link http://euklid.bauing.uni-weimar.de/ikm2009/paper.html 1611-4086 Klaus GürlebeckGND, Carsten KönkeORCiDGND English 2017/03/14 2010/07/14 2017/03/14 Bauhaus-Universität Weimar Bauhaus-Universität Weimar Fakultät Bauingenieurwesen / Graduiertenkolleg 1462 14 Computer Science Models in Engineering; Multiscale and Multiphysical Models; Scientific Computing Angewandte Informatik; Angewandte Mathematik; Architektur ; Computerunterstütztes Verfahren 000 Informatik, Informationswissenschaft, allgemeine Werke / 000 Informatik, Wissen, Systeme 500 Naturwissenschaften und Mathematik / 510 Mathematik 31 Mathematik / 31.80 Angewandte Mathematik 56 Bauwesen / 56.03 Methoden im Bauingenieurwesen Bauhaus-Universität Weimar / Internationales Kolloquium über Anwendungen der Informatik und Mathematik in Architektur und Bauwesen, IKM, Weimar / Internationales Kolloquium über Anwendungen der Informatik und Mathematik in Architektur und Bauwesen, IKM, Weimar, 18. 2009 Creative Commons 4.0 - Namensnennung-Nicht kommerziell (CC BY-NC 4.0)