Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
MODEL VALIDATION WITH CLASSICAL AND BAYESIAN HYPOTHESIS TESTING UNDER EPISTEMIC UNCERTAINTY.
Submission Author:
João Henrique Camargo , SP
Co-Authors:
João Henrique Camargo, Helio Fiori de Castro
Presenter: João Henrique Camargo
doi://10.26678/ABCM.COBEM2023.COB2023-1315
Abstract
As the systems used in engineering are increasing in complexity, physics-based simulations are being used to replace expensive physical tests for understanding system behavior. The computational model formulated to represent a physical system has an intrinsic uncertainty that needs to be accounted for. Model validation and verification is the processes of determining the degree to which a model is an accurate representation of the real world for its intended use. It involves comparison of model prediction with experimental data, expert intuition, theoretical results, or any combination of them. Epistemic uncertainty is expressed as a lack of knowledge of the physics, variations of the system and environment. A common approach to model validation is the graphical validation, a visually comparison between prediction and experimental observations, is still used to measure the accuracy of the model but this approach does not count for the uncertainty of the model and of system’s properties. In this paper a statistic-based quantitative approach will be used to account for these uncertainties, both in model prediction and experimental observation. The classical hypothesis testing is a well-developed approach method of rejecting or accepting a model based on an error statistic. The Bayesian hypothesis testing will compare two hypotheses based on the amount of information available, minimizing the risk of model selection by choosing properly the model acceptance threshold and an avoidance of making type I/II error. Two system responses will be analyzed: model analysis of a rotor test bench and free vibration of the beam. The uncertainty of the test bench are those intrinsic of the measurement, environment and epistemic uncertainty; the parameters of the beam model will be obtained by Monte Carlo Method to assure uncertainty. The expected results are the assessment of model validation and limits of acceptance for both analyses.
Keywords
model validation, Bayesian hypothesis testing, Hypothesis testing, Uncertainties, Uncertainties quantification
