Neural Network Methods Applicable to Predict the Noise Reduction Ability of Nonwoven Sandwich Absorbers

摘要

We propose to use general regression neural network (GRNN) to forecast the noise reduction ratio of a sandwich structure nonwoven absorber that bypasses the complex and heavy computation, which is more general compared with the models introduced in theoretical acoustics. The GRNN takes some easily measured structural parameters, such as thickness, area density, porosity, and pore size of each layer as inputs. The noise reduction ratio of each absorber is used as the GRNN's output. In experiment, one hundred sandwich structure nonwoven absorbers are particularly made of ten different types of meltblown polypropylene nonwoven materials and four types of hydroentangled E-glass fiber nonwovens initially. For comparison, the prediction model using back-propagation neural network is also built. The experiment results indicate that the prediction of noise reduction ratio using neural network-based method is reliable and efficient.