Title: Efficient Frequency Averaging Techniques for Noise and Vibration Simulations (Efficiënte technieken voor frequentiemiddeling in trillings- en geluidssimulaties)
Other Titles: Efficient Frequency Averaging Techniques for Noise and Vibration Simulations
Authors: D'Amico, Roberto
Issue Date: 27-May-2014
Abstract: In the latest decades, noise and vibration characteristics of products have been growing in importance, driven by market expectations and tightening regulations. Accordingly, CAE tools have become irreplaceable in assisting acousticians through the design process, and their accuracy and efficiency are essential to model the behavior of complex engineering systems.This dissertation aims at increasing the computational efficiency of deterministic simulation techniques for steady-state noise and vibration problems. In particular, the focus is on the efficient evaluation of weighted frequency integrals. Classic approaches make use of numerical quadrature to evaluate a frequency integral. However, the response of a vibrating system is commonly a highly oscillating function of frequency, and a large number of sampling points might be required to achieve an accurate integration. As an alternative, the residue theorem is proposed to compute the weighted integrals. The refined integration over real frequencies is replaced by a few computations at complex frequencies, with a consequent increased accuracy and computational efficiency.A weighted integral is first evaluated to compute the band-averaged input power into a vibrating system. The ideal rectangular weighting function is approximated by using the square magnitude of a Butterworth filter. Applying the residue theorem, the integral can be evaluated by computing the system response at a few points in the complex frequency plane. These points are some of the filter poles, equal in number to the order of the filter. This allows for an efficient integration, regardless of the bandwidth of analysis. Such a result is successively generalized. The band integral is computed by moving the path of integration to the complex frequency plane and applying efficient quadrature schemes. Due to the smoothness of the integrand in the complex frequency plane, the accuracy and efficiency of the technique are further increased. Moreover, it is shown that using numerical quadrature in the complex plane indirectly leads to the definition of a novel family of weighting functions over the real frequency domain.The proposed techniques for the evaluation of the band-averaged input power are accurate, efficient, easy to implement and can be employed within any classic deterministic framework. Applications allow assessing the effectiveness of the methodology for complex geometries and frequency dependent properties, and in combination with optimization schemes.When the order of the Butterworth filter is one, its square magnitude corresponds to the Lorentzian function, which allows computing the weighted average over a wide frequency range by evaluating the response at a single complex frequency. However, due to its bell-shape, the Lorentzian is not suitable to evaluate band values, and its use as a weighting function is investigated in three different ways. The ensemble mean input power is estimated by using the Lorentzian-weighted frequency averaging. The same procedure is used to evaluate the direct field dynamic stiffness of a component. Finally, the Lorentzian is used as a mass-frequency density function within the Fuzzy Structure Theory.
Publication status: published
KU Leuven publication type: TH
Appears in Collections:Production Engineering, Machine Design and Automation (PMA) Section

Files in This Item:
File Status SizeFormat
thesisDAmico.pdf Published 6830KbAdobe PDFView/Open Request a copy

These files are only available to some KU Leuven Association staff members


All items in Lirias are protected by copyright, with all rights reserved.