A Design Space Exploration Framework for Automotive Sound Packages in the Mid-Frequency Range

The continuous pursuit for lighter, more affordable and more silent cars, has pushed OEMs into optimizing the design of car components. The different panels surrounding the car interior cavity such as firewall, door or floor panels are of key importance to the NV performance. The design of the sound packages for high-frequency airborne input is well established. However, the design for the mid-frequency range is more difficult, because of the complex inputs involved, the lack of representative performance metrics and its high computational cost. In order to make early decisions for package design, performance maps based on the different design parameters are desired for mid-frequencies. This paper presents a framework to retrieve the response surface, from a numerical design space of finite-element frequency sweeps. This response surface describes the performance of a sound package against the different design variables. In order to build it, a surrogate model is used based on a Kriging method, which is iteratively enriched with new simulation points to increase the accuracy of the fitted model. Each new point consists of a finite element simulation in mid-frequencies, typically with porous layers based on the Biot formulation. In order 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. This paper applies the framework on a typical automotive package case under a structureborne excitation. Both material and geometrical parameters are considered to build a parametric response surface of the performance in the mid-frequency range. The application of this methodology helps to balance the mass, the cost and the performance of the silencers.