基于BP神经网络和响应面拟合的水工沥青混凝土动态抗压强度预测

Prediction of Dynamic Compressive Strength of Hydraulic Asphalt Concrete Based on BP Neural Network and Response Surface Fitting

为验证BP神经网络模型在预测水工沥青混凝土抗压强度方面的适用性,从应变率(10^(-5)~10^(-2) s^(-1))和温度(-20~30℃)2个维度设计了32组圆柱体试件进行单轴压缩试验。采用MATLAB软件建立BP神经网络模型,对水工沥青混凝土的抗压强度进行预测;采用麦夸特法(Levenberg-Marquardt)加通用全局优化法的优化算法,建立响应面拟合模型并与BP神经网络预测结果进行对比。结果表明,BP神经网络模型和响应面拟合模型预测值与试验值之间的相关系数r分别为1.0995和1.1142,相较于有具体表达式的响应面拟合模型,BP神经网络模型的预测精度更高,BP神经网络预测模型能够作为相关试验研究和数值分析的辅助手段。

To verify the applicability of back-propagation(BP)neural network models in predicting the compressive strength of hydraulic asphalt concrete,we designed 32 groups of cylinder specimens for uniaxial compression tests from the two dimensions of strain rate(10^(-5) s^(-1)~10^(-2) s^(-1))and temperature(-20~30℃).MATLAB was employed to establish a BP neural network model for the prediction of the compressive strength of hydraulic asphalt concrete.The response surface fitting model was established by the Levenberg-Marquardt method and the general global optimization algorithm,the prediction results of which were then compared with those of the BP neural network.The results show that the correlation coefficient r between the value predicted by the BP neural network model(the response surface fitting model)and the test value is 1.0995(1.1142).Compared with the response surface fitting model having specific expressions,the BP neural network model has higher prediction accuracy.The BP neural network prediction model can be used as an auxiliary means for related experimental research and numerical analysis.