Effect of model-form definition on uncertainty quantification in coupled models of mid-frequency range simulations. (1st September 2017)

• Record Type: Journal Article
• Authors: Van Buren, Kendra L.; Ouisse, Morvan; Cogan, Scott; Sadoulet-Reboul, Emeline; Maxit, Laurent
• Is Part Of: Mechanical systems and signal processing. Volume 93(2017)
• Page Start: 351
• Page End: 367
• Publication Date: 2017-09-01
• Subjects: Uncertainty quantification -- Statistical energy analysis -- Statistical modal energy distribution analysis -- Model reduction; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.02.020
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Uncertainty quantification of coupled numerical models is pursued. Analysis at the subsystem and coupled model levels is compared. Methodology applied to a four-subsystem model to predict mid-frequency response. Response is calculated using the Statistical Modal Energy Distribution Analysis. Abstract: In the development of numerical models, uncertainty quantification (UQ) can inform appropriate allocation of computational resources, often resulting in efficient analysis for activities such as model calibration and robust design. UQ can be especially beneficial for numerical models with significant computational expense, such as coupled models, which require several subsystem models to attain the performance of a more complex, inter-connected system. In the coupled model paradigm, UQ can be applied at either the subsystem model level or the coupled model level. When applied at the subsystem level, UQ is applied directly to the physical input parameters, which can be computationally expensive. In contrast, UQ at the coupled level may not be representative of the physical input parameters, but comes at the benefit of being computationally efficient to implement. To be physically meaningful, analysis at the coupled level requires information about how uncertainty is propagated through from the subsystem level. Herein, the proposed strategy is based on simulations performed at the subsystem level to inform a covariance matrix for UQ performed at the coupled level. The approach is applied to a four-subsystem model of mid-frequency vibrations simulated using the Statistical Modal Energy Distribution Analysis, a variant of the Statistical Energy Analysis. The proposed approach is computationally efficient to implement, while simultaneously capturing information from the subsystem level to ensure the analysis is physically meaningful.