广东省交通运输厅科技项目(科技 2016 02 004)
为准确预测苯乙烯丁二烯苯乙烯嵌段共聚物（SBS）改性沥青中SBS的含量，采用傅里叶变换红外光谱（FTIR）采集了不同SBS含量改性沥青的FTIR图谱，建立了基于深度神经网络（DNN）改性沥青中SBS含量的预测模型，并研究了不同因素对模型预测精度的影响，对比评价了模型的预测精度、敏感性及适用性.结果表明：数据的降维、降噪预处理使SBS含量预测模型的均方误差降低了70%；基于DNN改性沥青中SBS含量预测模型的精度高于标准曲线法和随机森林方法，其对改性沥青中SBS含量的预测具有较好的敏感性及适用性.
In order to determinate the content of styrene butadiene styrene block copolymer(SBS) in SBS modified asphalt accurately, the Fourier transform infrared spectroscopy(FTIR) spectra of modified asphalts containing different SBS contents were collected by using FTIR instrument, and the determination model for SBS content in modified asphalt was established based on deep neural network(DNN）. The influences of different factors on the accuracy of the determination model were studied, and the accuracy, susceptibility and applicability of the model were evaluated. The results show that mean square error of the SBS content determination model is reduced by 70% by dimension reduction and noise reduction. Determination accuracy for SBS content in modified asphalt using DNN method compares favourably with that using standard curve method and random forest method. It also has good sensitivity and applicability to determination of SBS content in modified asphalt by the DNN determination model.
王志祥,李建阁.基于DNN改性沥青中SBS含量的预测模型[J].建筑材料学报,2021,24(3):630-636
复制
