路面结构参数同时反算人工神经网络法

利用传递矩阵方法，计算出柔性路面结构在轴对称荷载作用及层间完全接触情况下的路表弯沉值，以该结果为输入数据训练四层BP网络，得到较精确的网络权值和阈值；在此基础上实现了对柔性路面结构层的模量和厚度参数的同时反算。计算分析表明，该方法速度快且精度高；即可反算路面结构层的模量，又能反算其厚度，通过与实际值比较，其计算精度完全满足工程实际的要求。

Artificial neural networks method for back-calculating pavement structural parameters simultaneously

By means of transfer matrix, the surface deflections of flexible pavement structure are calculated by the axisymmetrical loading case and with the boundary conditions of the layers being completely contacted. With the data, a BP network with four layers is trained. More accurate weight and threshold values for the BP network are obtained. Then, with these results, the modulus and thickness of flexible pavement structure are back calculated simultaneously. The comparison between calculation data and practical ones show that the results are satisfying. The accuracy and speed of this method are high. More important thing is that the method can back calculate both modulus and thickness of flexible pavement structure layers simultaneously.