摘要
基于小波去噪原理,对基坑变形数据小波去噪过程中的相关参数进行优化,对于去噪中的影响因素和作用规律研究有积极意义。利用最优小波去噪将原始监测数据分为趋势项序列和误差项序列,再利用BP神经网络对两序列加以预测,并与传统BP神经网络预测结果进行对比分析。结果表明:采用硬阈值取值和10层小波分解时的去噪效果最好,且通过去噪分离了原始数据的长期性和游离性,增加了数据的可预测性;并且由后期预测结果可知,小波神经网络预测精度要优于传统的BP神经网络预测,具有更高的可信度。
Based on the theory of wavelet de-noising, relevant parameters in the wavelet de-nosing process of foundation pit deformation were studied, which is helpful for the study of influential factors and laws in de-noising. The original monitoring da-ta were classified as trend sequence and error sequence, and the two sequence data were predicted by BP neural network. The prediction results were also compared with the analysis results by traditional BP neural network. The results show that using hard threshold value and 10-layer wavelet decomposition can obtain the best results in de-noising process. The data predictability is increased by data de-noising that separates the long-term and free features of the original data. The subsequent forecast result shows that the forecasting precision of wavelet neural network method in this paper is superior to the traditional BP neural network method.
出处
《人民长江》
北大核心
2014年第19期41-46,共6页
Yangtze River
关键词
基坑变形
小波去噪
R/S分析
BP神经网络
foundation pit deformation
wavelet de-noising
R/S analysis
BP neural network
