Back to Search
Start Over
Virtual reality of interior noises of vehicles based on boundary element and neural networks.
- Source :
- Neural Computing & Applications; Mar2018, Vol. 29 Issue 5, p1281-1291, 11p
- Publication Year :
- 2018
-
Abstract
- Interior noises of vehicles would be caused when the vibration of body panels excites the indoor air. In the paper, the vibration load of engine was obtained firstly through experiments. Secondly, the engine load was applied in the finite element model of body in white to compute the vibration velocity and realize virtual reality, indicating that the front support of the body had large vibration velocity when the frequency was lower than 60 Hz. The boundary element was then adopted to compute the interior noise and extract the sound pressure at a point near the driver’s head. Two obvious peaks were shown in sound pressure curves, at 270 and 310 Hz, respectively. The body panels that had obvious impact on the interior peak noise were determined through the panel contribution analysis, and the interior peak noise was remarkably reduced after applying sound absorption materials on these panels. Nevertheless, many more additional sound absorption materials were not always better. If a multilayer of sound absorption materials was needed, an optimal value was existed in the thickness of sound absorption material. And a great impact would be played toward the interior noise of the cabin by the reasonable selection of different sound absorption materials and their thicknesses. Finally, the neutral network (NN) was also used to predict interior noises, which was compared with the result of the boundary element. The maximum difference between the prediction values of NN and boundary element was within 5 dB, indicating that the neural network was feasible to predict the interior noise. Subsequently, the neural network method would be applied to conduct the optimization analysis for the interior noise. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 09410643
- Volume :
- 29
- Issue :
- 5
- Database :
- Complementary Index
- Journal :
- Neural Computing & Applications
- Publication Type :
- Academic Journal
- Accession number :
- 128333193
- Full Text :
- https://doi.org/10.1007/s00521-016-2836-0