灰色模型结合神经网络预测高速公路路基沉降

Grey model combines neural network to predict highway sub-grade settlement

为了提高高速公路路基沉降的预测精度,考虑到神经网络强大的非线性映射功能,提出了"灰色模型+神经网络"对高速公路路基沉降进行预测分析的组合方法。以湖南省某高速公路路基沉降多个断面实测数据构建灰色GM(1,1)预测模型,在采用构建的灰色模型预测出相应结果的基础上,运用神经网络对预测结果做误差补偿。研究结果表明,采用实测数据拟合的灰色模型预测值的最大相对误差与运用神经网络对预测结果做误差补偿之后的优化预测值的最大相对误差分别为19.193%和0.865%,用神经网络对灰色模型预测结果做误差补偿之后的优化预测值与实测值更接近。

In order to improve the accuracy of highway sub-grade prediction,taking into account the neural network powerful nonlinear gray model＋neural network＂ was offered to mapping function, a combined method of ＂ predict highway foundation settlement, based on the GM（1,1） model to predict the result in the application of neural network to compensate its error. Taking the case of highway sub-grade settlement of Hunan province to build predictive models, the results show that the maximum relative error of gray model prediction with neural network optimization gray model predictions were 19. 193% and 0. 865%, after optimization model predictions of gray to use neural networks is closer effective records.