基于ANN算法与分形理论的沥青混合料蠕变预测

Prediction of Creep Deformation of Asphalt Concrete Based on Ann Algorithm and Fractal Theory

目的研究AC-10、AC-13和AC-6沥青混合料的分形维数、级配等物理力学参数,及其对蠕变变形的影响.提供一种新的沥青混凝土路面变形预测与控制的研究方法.方法针对沥青混合料颗粒级配的自相似与长期稳定变形多因素性的特征,分别采用级配筛分实验与CT扫描数字图像技术对自相似特征进行验证与分形维数的获取,利用TAW-2000岩石三轴实验仪对AC-10、AC-13和AC16密级配沥青混合料开展三轴压缩实验与蠕变实验,基于人工神经网络算法(ANN)对实验数据进行驯化,并对蠕变实验的长期变形量进行预测.结果沥青混合料级配具有高度的自相似特征,分形维数在2.45~2.50.结论人工神经网络建立的长期变形预测模型,可以较好地对一定应力水平下的长期蠕变变形进行预测.

To provide a new research method of deformation prediction and control of asphalt concrete pavement,the physical and mechanical parameters of AC-10,AC-13 and ac-6 asphalt mixtures,such as fractal dimension and gradation,and their influence on creep deformation are studied.According to the characteristics of self similarity and long-term stable deformation of asphalt mixture particle gradation,gradation screening experiment and CT scanning digital image are used to verification of the self similar characteristics and obtain of the fractal dimension.Triaxial compression experiment and creep experiment are carried out on AC-10,AC-13 and ac16 dense graded asphalt mixture by the TAW-2000 rock triaxial test instrument.The experimental data are domesticated based on the artificial neural network algorithm(ANN),and the long-term deformation of creep experiment is predicted.The gradation of asphalt mixture has a high self similarity.The fractal dimension is 2.45~2.5.The long-term creep deformation prediction model based on artificial neural network can predict the long-term creep deformation at a certain stress level.