Accuracy of vibro-acoustic computations using non-equidistant frequency spacing

Within the vibro-acoustic optimisation of complex components under dynamic loading the radiated sound power is commonly used as an objective. For this purpose, the frequency-dependent sound power has to be quantified by a single scalar objective. For the required steady state simulations the mode-based frequency spacing can include non-equidistant step sizes as well as can change due to structural or material modifications. Thus, the total number of frequency steps is depending on the number of contributing modes that can be changed during optimisation processes with structural or material modifications. Furthermore, the accuracy of the objective has to be assured by choosing the required number of frequency steps and avoiding either under-resolved peaks or too many frequency steps. In this study, we present an approach for the determination of the averaged sound power within the covered frequency range with non-equidistant spacing based on the power spectral density. These scalar quantities are robust to any model changes. Thereafter, the mean power is used as a convergence criterion to determine the number of required frequency steps for a single mode and thus to reduce the computational efforts to a minimum. Further, a recommendation for a common rule for the spacing of single mode is given. This results in the frequency spacing estimation depending on the distance of neighbouring modes as well as the damping and biasing. Moreover, the combination of robust scalar objectives and efficient frequency spacing opens the prospects of accessing sound power objectives for complex optimisation problems. (C) 2018 The Authors. Published by Elsevier Ltd.