Metamodel-based sensitivity analysis and optimization of transmission loss in roller shutter boxes

Abstract

This study presents a detailed investigation of the acoustic performance of roller shutter boxes in the extended position, comprising porous and heavy mass layers. Such configurations are of particular relevance, as roller shutter boxes often represent weak points in the façade sound insulation of residential and office buildings. The Finite Transfer Matrix Method (FTMM) is first employed to model the system, providing a robust description of its acoustic properties, with numerical predictions validated against laboratory measurements. A two-step global sensitivity analysis is then performed to identify the dominant parameters. The Morris method is initially applied to screen mechanical, acoustic, and geometric variables, followed by the computation of Sobol indices to quantify the influence and interactions of the retained parameters. Given the high computational cost of Sobol analysis for complex models, several metamodeling techniques are evaluated, including Polynomial Chaos Expansion, Kriging, and Polynomial Chaos Kriging (PCK). Among these, PCK is shown to provide the most accurate and efficient framework for estimating sensitivity indices. Finally, an optimization procedure based on a genetic algorithm is conducted on the micro-macro model of polyurethane foams, focusing on controllable material parameters. The results emphasize the critical role of the reticulation rate in enhancing the sound transmission loss of extended roller shutter boxes.