Comparison of Clustering-Based Virtual SEA Subsystem Generation Models

Abstract

Building a valid Statistical Energy Analysis (SEA) model is far from straightforward in real life applications because it has to comply with certain requirements. Since intuition-based SEA subsystems do not always entirely meet these, more advanced methods such as clustering techniques have been proposed for model building. This paper focuses on the comparison of different clustering techniques from SEA model validity point of view in the case of a simplified car body model. Besides the compliance with the SEA theory, each obtained model is compared to a reference finite element simulation to assess its accuracy.