基于改进BP神经网络的软土路基沉降预测研究
Settlement Prediction of Soft Soil Roadbed Based on Improved BP Neural Network
王恒 1王佼佼1
作者信息
- 1. 中冶南方城市建设工程技术有限公司,武汉 430070
- 折叠
摘要
结合软土路基沉降观测值,分别从精度、安全性和预测期限等角度对双曲线模型、S型成长曲线模型和BP神经网络进行适用性分析.为了提高沉降预测精度、克服传统BP神经网络缺陷,采用LM优化算法改进BP神经网络,验证了 LM算法的优越性,在此基础上建立了一种考虑趋势部分和随机部分的组合预测模型.该模型既符合路基沉降的发展规律,又能够充分利用BP神经网络的非线性外推能力,弥补趋势函数的精度下降.工程实例研究表明,该模型优于各单一预测模型,具有很好的预测效果,均方差为0.03,残差平方和为1.8,相关系数接近1.
Abstract
The applicability of some prediction models,such as hyperbolic model,S type growth curve model and BP neural network,is analyzed from the view of accuracy,safety and prediction term according to the settlement observation values of soft soil roadbed.In order to overcome the shortcomings of the traditional BP neural network and increase the prediction accuracy,LM optimization algorithm is used to improve the BP neural network.The superior of LM algorithm is verified,and a combination forecasting model taking the trend and the random parts into account is established.This model not only accords with the development rule of roadbed settlement,but also can make full use of the nonlinear extrapolation ability of BP neural network,which can make up the decline of the trend function.The engineering example shows that the model is superior to the single prediction model,and has better prediction effects:the MSE is only 0.03,the SSE is only 1.8,the correlation coefficient is almost 1.
关键词
沉降预测/BP神经网络/组合预测/LM算法/适用性Key words
Settlement prediction/BP neural network/Combination forecast/LM optimization algorithm/Applicability引用本文复制引用
出版年
2024
NETL
NSTL
万方数据