摘要
为了克服传统水泥路面使用性能预测方法的缺陷和误差反向传播(BP)神经网络的不足,利用动量方法改进了BP神经网络收敛性,建立了水泥路面使用性能预测模型.采用广东水泥路面调查数据对模型进行了训练和验证,并对模型训练方法进行了优化.分析表明,该模型具有较好的实用性和预测精度.
In order to deal with the deficiency of traditional prediction method of pavement penormance and the insufficiency of Back-Propagation（BP） neural network, a prediction model based on the improved neural network with momentum Back-Propagation （MOBP） is developed. The model is validated and trained with pavement performance data of Guangdong province, and the training method is also optimized. According to the theoretical analysis and practical verification, the approach is completely feasible.
出处
《同济大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2006年第9期1191-1195,共5页
Journal of Tongji University:Natural Science
关键词
水泥混凝土路面
路面使用性能
神经网络
预测模型
portland cement concrete pavement
pavement performance
neural network
prediction model
