基于优化神经网络的沥青混合料力学特性预估

Prediction of mechanical properties of asphalt mixtures based on optimized neural networks

现有的沥青混合料疲劳寿命预估大多基于传统的疲劳方程拟合得到,但由于路面结构的多向性和材料的复杂性,其预测精度往往不尽人意。为了提高预测精度,在遗传算法的基础上对神经网络架构进行优化,通过室内间接拉伸试验建立了沥青混合料强度及疲劳寿命预估模型,并对预估模型的精度进行了验证。试验结果表明,采用遗传算法优化的神经网络用于预测沥青混合料疲劳力学特性精度误差在4%以内,远优于传统的疲劳预测方程,可以作为获取沥青混合料疲劳特性研究数据的一种有效方法。

The existing fatigue life prediction of asphalt mixtures is mostly based on traditional fatigue equation fitting;however,due to the multidirectionality of pavement structure and the complexity of materials,the prediction accuracy is often not satisfactory.Therefore,this article establishes an optimized neural network-based model for predicting the strength and fatigue life of asphalt mixtures using indoor indirect tensile tests and verifies the accuracy of the prediction model.The experimental results show that the accuracy of Genetic Algorithm-Back Propagation neural network to predict the fatigue mechanical properties for asphalt mixture is within 4%,which is far superior to traditional fatigue prediction equations and can be used as an effective method to obtain data on the fatigue characteristics of asphalt mixtures.