基于GA优化灰色神经网络冷再生材料力学特性预测分析

Prediction and Analysis of Cold Recycled Materials Mechanical Properties based on Genetic Algorithm Optimized Grey Neural Network

通过室内试验测得不同路面结构下冷再生材料力学特性数据,运用遗传算法(GA)优化灰色神经网络模型进行数据分析和预测,采用灰关联理论进行冷再生材料力学特性影响因素(冷再生层厚度和模量、水泥稳定碎石厚度和模量以及土基模量)的敏感性分析。结果表明,运用遗传算法(GA)优化灰色神经网络组合预测值与试验实测值最大误差仅为6.281%,能有效预测乳化沥青冷再生材料力学特性,可对不同因素下冷再生力学特性进行量化预测分析,可减少试验量。通过灰关联理论敏感性分析得到,水泥稳定碎石模量对乳化沥青力学性能影响较大。

Through indoor test the mechanical properties data of cold recycled materials are collected under different pavement structure, and the grey neural network combination model optimized by genetic algorithm （GA） is used for data analysis and forecasting, by means of grey relational analysis method to analyze influencing factors （thickness and modulus of emulsified asphalt cold regeneration layer and the cement stable macadam mixture, and modulus of soil base）. Results indicate that the maximum error is only 6. 281% between the predicted results by using the grey neural network combination model optimized by genetic algorithm （GA） and actual measured results, which can effectively predict the mechanical properties of emulsified asphalt cold recycled materials and also can reduce quantities of experiment. By means of grey relational analysis method, it shows that modulus of cement stabilized macadam has great influence on the mechanical properties of cold regeneration material.