基于BP神经网络与深度学习模型的旧水泥混凝土板共振碎石化层动态模量分析

Dynamic Modulus Analysis of Resonance Rubblization Layer of Old Cement Concrete Slab based on BP Neural Network and Deep Learning Model

为研究旧水泥混凝土路面板共振破碎后的碎石化层在动态荷载作用下的承载性能与结构层动态模量,首先采用落锤式弯沉仪(FWD)现场检测采集共振碎石化加铺沥青混凝土面层后的路表弯沉盆,然后基于深度学习模型与BP神经网络训练获得的3层沥青路面结构模量反算模型计算各结构层模量,在此基础上通过动态有限元建模对比分析路表计算弯沉盆与FWD实测弯沉盆。研究结果表明:基于BP神经网络与深度学习模型开发的动态模量反算模型具有较高的可靠度和准确性,能根据FWD实测的路表弯沉盆反算旧水泥混凝土路面板共振碎石化加铺沥青路面的各结构层动态模量。旧水泥混凝土板共振碎石化层动态模量远高于普通级配碎石基层,表明共振碎石化层具有优异的承载能力。

In order to study the bearing capacity of the rubble layer after the resonance crushing of the old cement concrete pavement slab under dynamic load and the dynamic modulus of the structural layer,firstly,the falling weight deflectometer(FWD)is used to detect and collect the road surface deflection basin after the resonance rubble paving asphalt concrete surface layer.Then,based on the deep learning model and the three-layer asphalt pavement structure modulus inverse calculation model obtained by BP neural network training,the modulus of each structural layer is calculated.On this basis,the dynamic finite element modeling is used to compare and analyze the calculated deflection basin of the road surface and the measured deflection basin of FWD.The research results indicate that the dynamic modulus backcalculation model developed based on BP neural network and deep learning model has high reliability and accuracy,and can backcalculate the dynamic modulus of each structural layer of the old cement concrete pavement slab resonant rubblization overlay asphalt pavement according to the surface deflection basin measured by FWD.The dynamic modulus of the resonant rubblized layer of the old cement concrete slab is much higher than that of the ordinary graded gravel structure layer,indicating that the resonant rubblized layer has excellent bearing capacity.