Bottom-up calibration of an industrial pump model: Toward a robust calibration paradigm

Abstract

The calibration of complex industrial structures using vibration test data is an important step toward improving the credibility of model-based decisions for a specified application. Meanwhile, the state-of-practice in industry is to use computer-aided design software that produce very refined finite element meshes thus creating an impractical bottleneck in the iterative calibration process. This paper illustrates a two level bottom-up approach that requires testing at both the component and assembled levels. Initially, a global sensitivity analysis is performed on the complete model to rank the model components in terms of their influence on the quantities of interest. Selected components are then calibrated using dedicated tests before being integrated as a Craig-Bampton superelement into the global assembly. At the top level, model calibration is restricted mainly to the component interface properties. Since test data is available only for a single pump, a deterministic calibration paradigm is applied here. This two level procedure is illustrated on a detailed model of a pump that is studied in the framework of the French nationally funded project SICODYN. A robust calibration methodology will also be outlined for future work in order to account for lack of knowledge in the final operational boundary conditions of the pump.