基于神经网络和双曲线混合模型的高速公路沉降预测

Highway subsidence prediction based on neural network and hyperbolic hybrid model

为了提高高速公路沉降预测的精度,提出了"双曲线+神经网络"的混合模型对高速公路沉降进行预测.基于某省某高速公路K57+580M断面2003—2013年的实测沉降数据,采用双曲线模型对施工期和通车期分别进行沉降预测.然后利用构建的混合模型重新对该断面进行施工期和通车期的沉降预测,分析比较2种模型的预测精度.为了验证2种模型的预测精度,选取该高速公路K156+100M断面实测数据进行计算.结果表明,混合模型在施工期和通车期的预测值的中误差分别达到1.13和7.30 mm,预测精度相比双曲线模型在施工期和通车期分别提高了66.2%和41.8%.

In order to improve the accuracy of subsidence prediction, a hybrid model based on neural network and hyperbolic is proposed for highway subsidence prediction. The used data is the actual observed settlement for the K57 ＋ 580M section in a highway project of one province from 2003 to 2013. The data is calculated by using the hyperbolic model and the hybrid model in the construction