Simulating Acoustic Engine Performance Over a Broad Frequency Range 2011-26-0019
Acoustic performance of vehicle engines is a real challenge for powertrain design engineers. Quiet engines are required to reduce noise pollution and satisfy pass-by noise regulations, but also to improve the driving comfort. Simulation techniques such as the Boundary Element Method (BEM) have already been available for some time and allow predicting the vibro-acoustic response of engines. Although the accuracy of these simulation techniques has been proven, a challenge still remains in the required computation time. Given the large amount of speeds for a full engine run-up and the need to cover a large frequency range, computation times are significant, which limits the possibility to perform many design iterations to optimize the system.
In 2001, Acoustic Transfer Vectors (ATV)  have been presented to adequately deal with multiple rpm. The ATV provide the acoustic response for unit surface velocities and are therefore independent from the engine's actual surface vibrations. As such, the ATV only need to be computed once and can be easily combined afterwards with the actual vibrations at each rpm to obtain the acoustic response for a full engine run-up.
This paper presents recent further improvements to reduce the computation time of engine acoustic (ATV) simulation. For BEM, the Fast Multipole BEM method is discussed. For Finite Element Methods (FEM), a new modeling approach based on the Perfectly Matched Layer (PML) technique is presented and an update on state of the art iterative (Krylov) solvers and direct solvers is provided. This paper concludes with a discussion on the results of two industrial simulation cases in which these new techniques have been applied.