Study On The Sound Absorption Property For Multi-layer Composite Nonwovens And The Prediction By Using Neutral Network Model

With the development of urbanization, there are more and more noise pollution in our life, how to decrease the damage that noise takes has been a common concern of researchers around world nowadays. Nonwovens is a kind of typical porous materials with the advantage of light weight, high intensity, and low costs, and has been the main development trend of new materials used for sound absorption. In this paper, we research the sound absorbing property of composites made of melt-blown nonwovens and glass fiber materials.By gathering the melt-blown nonwovens and glass fibers to be A+B or A+B+A’ model and by testing composites’ thickness、surface density、bulk density、porosity、 aperture and sound absorption coefficient, we made research to the influence factor of composites’ sound absorption property. What’s more, we predicted the composites’ average sound absorption coefficient by BP and GRNN neutral network to calculate the deviation between forecast result and test result.The experiment shows that, with the increasing of the thickness of melt-blown nonwovens and glass fiber materials, the sound absorption band became wider, its coefficient peak moved to low frequency direction and sound absorption effect improved evidently. Besides, the arrangement of material porosity had great effect on the composites’ sound absorption property. We found that establishing sound absorption model by BP neutral network could complete the forecasting task but it still had some flawless:one is that the choice of number of the neurons in hidden layer of the BP neural network has no specific theory;the other is that the shortage of data in training may lead larger deviation between forecast result and expected results. GRNN neutral network has dominant position compared with BP neutral network when it was used to sound absorption coefficient forecasting as it has less deviation and better stability.