基于人工神经网络的沥青路面模量评估

Artificial Neural Network-Based Evaluation of Layer Modulus of Asphalt Pavement

介绍了利用弯沉盆参数和人工神经网络方法来处理FWD(落锤弯沉仪)的测量结果用于结构层模量的评估。在研究弯沉盆信息时,利用FORTRAN程序进行了轴对称模型的动力有限元分析.与现有的反算程序中利用迭代不断调整模量来匹配弯沉的方法不同,本方法利用基层破坏指数和形状因子两个参数来确定土基模量,并将土基模量和其他参数一起作为输入,利用人工神经网络来预估上层模量.最后,分别利用本方法和软件ELMOD对实测FWD数据进行分析,结果验证了本方法的可行性.

Based on deflection basin parameters and artificial neural networks （ANNs）, this paper presents a methodology for processing dynamic falling weight deflectometer （FWD） measurements to estimate layer module and condition. A two-dimensional dynamic finite element analysis is made of the deflection information by using FORTRPuN. Unlike the majority of the existing backcalculation programs that iteratively adjust all the layer module to match the measured deflections, the proposed method first determines the subgrade modulus by means of two deflection basin parameters, namely Base Damage Index and Shape Factor F2 ; and then the estimated subgrade modulus and other parameters as input variables are applied to a trained ANN to estimate the upper layers＇ module. Field FWD measurements are analyzed by this method and ELMOD4 program respectively, the results of which validates the feasibility of the proposed method.