摘要
为了提高路基弯沉检测的效率和精度,以便携式落锤弯沉仪PFWD为平台,对路基弯沉盆指标进行了分析研究.并且针对含砾黏土和含碎石黏土2种不同路基结构,基于回归分析技术及人工神经网络智能方法构建了动弯沉与贝克曼梁静态弯沉之间的分析模型,并对2种不同模型的预测结果进行评价分析.实测数据表明,由于考虑了弯沉盆指标,采用回归模型对2种不同路基的贝克曼梁静弯沉预测相对误差均值分别由4.64%和3.99%下降到3.01%和2.35%,回归方程对贝克曼梁静弯沉的预测精度得到了提高;同时,采用BP神经网络模型预测相对误差均值分别为1.66%和1.80%,优于多元回归模型.研究结果可以为路基贝克曼梁静弯沉的检测提供参考.
In order to improve the efficiency and precision of subgrade deflection detection, this pa- per analyzes subgrade deflection basin indicators using portable falling weight deflectometer (PF- WD) as the platform. For different embankment structures, regression analysis and artificial neural network method are used to establish models relating dynamic deflection to the static deflection. The measured data show that, considering the deflection basin indicators, the average of the relative error of Benkelman beam deflection predicted by regression models are reduced from 4.64% and 3.99% to 3.01% and 2.35% for two different embankment structures respectively. Meanwhile, the average of the relative error predicted by BP (back-propagation) neural network model are 1.66% and 1. 80%, which is better than the multiple regression models. The results provide reference for predic- tion of static deflection.
出处
《东南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2012年第5期970-974,共5页
Journal of Southeast University:Natural Science Edition
基金
教育部"长江学者"特聘教授奖励基金资助项目
美国国家科学基金资助项目(CMMI-0644552)
江苏省"六大人才高峰"资助项目
教育部霍英东基金资助项目(114024)
江苏省自然科学重点资助项目(SBK200910046)
关键词
弯沉
携式落锤弯沉仪
贝克曼梁
神经网络
deflection
portable falling weight deflectometer
Benkelman beam
artificial neural network