基于主成分分析和神经网络的马歇尔试验模型

Marshall Test Model based on Principal Component Analysis and Neural Network

为解决沥青混合料马歇尔试验存在的试验周期长、材料浪费等问题,文中基于主成分分析(PAC)、BP和RBF神经网络算法提出了一种沥青混合料马歇尔试验模型。首先,使用PCA对沥青混合料的合成级配参数和油石比进行变量降维,得到5个主成分;然后,将5个主成分作为神经网络模型的输入变量,稳定度和流值作为模型的输出变量,对模型进行训练;最后,将模型输出结果与实验室试验结果进行对比,验证模型的有效性。结果表明,BP神经网络对稳定度和流值输出的平均相对误差为5.19%和2.61%,RBF神经网络为4.95%和6.67%;BP和RBF神经网络运行时间分别为0.557s和0.962s。

In order to solve the problems of long test period and material waste in asphalt mixture Marshall test,in the paper,a Marshall test model of asphalt mixture is proposed based on principal component analysis(PAC),BP and RBF neural network algorithm.First,the PCA is used to reduce the dimensionality of the synthetic gradation parameters and the oil-stone ratio of the asphalt mixture to obtain five principal components.Then,the five principal components are used as the input variables of the neural network model,and the stability and flow values are used as the output variables of the model to train the model.Finally,the model output is compared with the laboratory test results to verify the validity of the model.The results show that the average relative error of BP neural network for stability and flow value output is 5.19%and 2.61%,RBF neural network is 4.95%and 6.67%;BP and RBF neural network running time are 0.557 sand 0.962 srespectively.