基于BP神经网络的非织造材料基复合吸声体吸声系数预测

Acoustic absorption coefficient prediction of nonwovens matrix composite based on BP neuron network

以涤纶针刺非织造材料和聚丙烯熔喷非织造材料为研究对象,通过实验获得其物理结构参数,并将复合前后非织造材料厚度、面密度、孔隙率和孔径作为BP神经网络的输入项,用于预测吸声体的平均吸声系数,同时通过调节输入神经元个数、传递函数和隐含层个数构建了最佳的BP神经网络预测模型。对非织造材料基复合吸声体的吸声性能进行预测,并与测试结果进行了对比。结果表明,运用BP神经网络可以建立较理想的适用于复合吸声体平均吸声系数预测的模型。

Polypropylene melt-blown nonwovens and polyester fiber needle-punched nonwovens as the research object, through experiments its physical structure parameters can be obtained. The thickness, density, pore size and porosity of composite nonwovens at composition before and after as the input of Back Propagation （ BP） neural network were used to predict the average sound absorption coefficient of the nonwovens matrix composite absorber. In order to get the optimum network, the number of input neuron, transfer function and the number of hider layer were adjusted. The average sound absorption coefficient of the nonwovens matrix composite absorber was predicted by BP neural network, and was compared the predicting outcomes with test result. The result showed that with the optimum BP neural network can be set up ideal model for nonwovens matrix composite absorber.