改进的BP神经网络在路基沉降预测中的运用

The Application of the Improved BP Neural Network to the Subgrade Settlement Prediction

针对BP神经网络存在训练过程不确定的缺点,基于MATLAB建立改进的BP神经网络模型,该模型可克服BP神经网络模型在训练过程中收敛速度慢、易陷入局部极小点等缺点。结合实体工程实测数据,将该优化模型与指数曲线模型、双曲线模型、泊松曲线模型和Compertz模型对比分析,结果表明改进的BP神经网络模型在黄土路基沉降预测中精度最高,可运用于黄土路基的沉降预测。

Combined with the real measured data of the entity engineering, this paper compared and analyzed the optimization model, exponential curve model, hyperbolic model, Poisson curve model and Compertz model. The results showed that the improved BP neural network model accuracy was the highest in the loess sugrade settlement prediction, which could be used for the settlement prediction of loess subgrade.