Identification and Quantification of Multivariate Possibilistic Manufacturing Uncertainty in Dynamic Numerical Models

Abstract

Abstract. The objective of this work to validate a novel methodology for the identification and quantification of possiblistic multivariate uncertainty that has been presented by the authors in previous work. The method is based on the convex hull concept for both the representation of uncertainty in the result of an interval finite element computation, as for the variability that was measured on the real-life structure. Identification of the parameteric multivariate interval uncertainty is performed by minimisation of a metric describing the discrepancy between these convex hulls. This method has been proven to be able do deliver an accurate identification of multivariate possibilistic uncertainty, which is also robust against certain measurement set metrics, however only on small-scale academic examples. This paper therefore first introduces a generic method for the reduction of the dimensionality of the identification at hand, and shows a validation of the method on a high-dimensional, complicated numerical model. Specifically, a test structure containing uncertainty in the stiffness of several bolted connections, introduced during the assembly of the structure, will be considered. The uncertainty in the stiffness of these connections is identified by using the presented method.