摘要
根据沉降数据的特性,以最小二乘支持向量机为核心技术构建预测模型,提出了一种路基沉降预测的新方法。由于测量误差不可避免,沉降数据通常含有噪声,不宜直接进行拟合,因此首先采用小波分析的方法对原始沉降数据进行降噪预处理,然后馈送到最小二乘支持向量机完成沉降预测。最后用某高速公路实测数据进行了实例分析,并与BP神经网络预测结果进行了对比,计算结果表明,小波分析结合支持向量机的模型有较好的预测精度,将该模型应用于公路软基沉降预测是可行的和值得研究的。
A new model for foundation settlement prediction was proposed. Considering the characteristics of settlement data, least squares support vector machine （LSSVM） was adopted as kernel technique to build a settlement prediction system. Settlement data usually have noises because the measurement error is inevitable and it is irrational to use the initial data directly in the model. The initial settlement data was denoised by wavelet analysis, then they were sent to the LSSVM for settlement prediction. Finally, the model was tested using the field data of an expressway. The comparison between SVM and ANN method was made. The forecasted results show that this model, which combined wavelet and LSSVM has well forecasting performance, is feasible to forecast settlement of road soft foundations.
出处
《公路交通科技》
CAS
CSCD
北大核心
2008年第6期40-43,49,共5页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金重点资助项目(50639010)
