Design of automotive sound packages in the mid-frequency range

The pursuit for green, affordable and comfortable cars has pushed the automobile manufacturers to balance the mass, the cost and the performance of every component. This is notably valid for sound packages, which are key components to the interior noise and vibration performance. Their design for the low- and high-frequency range is well established. However, the design for the mid-frequency range (100-1000 Hz) is more difficult, because of the mixed contributions of airborne and structureborne noises, the lack of representative performance metrics and the high computational cost of the simulations. Because of those difficulties, the design process of sound packages in the mid-frequency range is not streamlined. Eventually, the performance for the mid-frequency range mostly relies on the evaluation on an actual vehicle, which is available at a late stage, when the freedom to modify the design is low.

To develop the sound packages for the mid-frequency range, a framework has been implemented. It consists in a V-shape process to breakdown the interior noise and vibration performance targets to the inputs and body, and finally to the sound package design. This dissertation focuses on the development of tools and guidelines to support the design of the sound packages.

Performance metrics for sound packages in the mid-frequency range are investigated. Two transmission metrics are selected: the first one quantifies the losses in terms of velocity transmission through the panel and the second through the silencer. Additionally, a structureborne back-coupling metric is developed to quantify how the sound package accepts a certain structureborne input.

A framework to compute performance maps of sound packages is proposed. This framework embeds a model order reduction method to speed up the simulations and a surrogate modelling method to reduce the number of designs. The surrogate model is used based on a Kriging method, which is iteratively enriched with new designs to increase the accuracy of the fitted model. Each new design consists of a finite element simulation of a sound package in the mid-frequency range, with porous layers based on the Biot formulation. To gain computation time, it is run using a reduced order model based on a Krylov projection matrix-free algorithm: from a limited set of frequency lines, each metric is approximated by a reduced model using rational functions. Through an iterative enrichment, new frequency lines are added until sufficient accuracy is achieved on all the metrics.

An experimental validation of this framework is carried out. 47 specifications of sound packages are tested on a structureborne set-up and their numerical counterpart is simulated by the finite element method. These explicit simulations are correlated to the measurements in local transfer function in values and in main effects of two design parameters.

A structureborne application of this framework on a typical sound package is presented. Both material and geometrical parameters are considered to build a parametric response surface of the performance in the mid-frequency range. From a limited number of designs, the framework could build surrogate models representing the three metrics. Two types of guidelines are extracted: (1) guidelines to get the best performance under mass and height constraints; (2) guidelines in terms of ranges and slopes for each parameter. The robustness of these guidelines is assessed.