【摘要】 基于传热学原理初步确定沥青路面高温温度场影响因素，然后利用SPSS软件对沥青路面温度与影响因素之间相关性进行分析，根据分析结果选择与沥青路面温度具有较高相关性因素作为输入变量，最后分别采用长短期记忆网络LSTM(Long Short-Term Memory)、BP神经网络和传统回归方法，建立高温期沥青路面温度预测模型。结果表明，气温、太阳辐射强度、相对湿度、地面气压、路面深度与路面温度具有较高相关性；相较于传统回归方法和BP神经网络建立模型，基于LSTM神经网络建立的温度预估模型精度评价指标均较优，且预测温度曲线与实测温度曲线吻合程度最高，表明LSTM神经网络方法适用于沥青路面高温温度场预估。更多还原

【Abstract】 Based on the principle of heat transfer, the factors influencing the temperature field of asphalt pavement were preliminarily determined. Then, the correlation between the temperature of asphalt pavement and the influencing factors was analyzed by using SPSS. According to the correlation analysis results, high correlation factors with the temperature of asphalt pavement were selected as the input variables.Finally, LSTM(Long Short-Term Memory) neural network, BP neural network and traditional regression were used to establish the prediction model of asphalt pavement temperature field in high temperature period.The result shows that the air temperature, solar radiation intensity, relative humidity, surface air pressure, road depth are highly correlated with pavement temperature.Compared with the model established by traditional regression method and BP neural network, the best precision evaluation indicators are acquired from the model established by LSTM neural network.The predicted temperature curve from LSTM neural network is in well agreement with the measured temperature, indicating that LSTM neural network method is suitable for the prediction of asphalt pavement high temperature field.更多还原