基于经验模态分解方法的非稳定性基坑变形研究

Unstable foundation pit deformation based on empirical mode decomposition method

基坑施工是众多工程建设的基础和前提,科学预测其变形特征是一项十分重要的工作。本文针对神经网络模型在非稳定性基坑变形预测方面的研究现状,以进一步提升预测精度为导向,借鉴非线性非平稳信号处理的思路提出了基于经验模态分解的非稳定性基坑预测方法。该方法首先将非稳定变形数据分解为具有一定变化规律的子数据集,然后采用粒子群优化算法-反向传播(Particle Swarm Optimization-Back Propagation,PSO-BP)神经网络模型对子数据集进行变形预测,最后将各子数据集的预测结果叠加后即可得出可靠的预测结论。试验结果表明,该方法的预测精度显著提高,均方根误差仅为0.417 mm,平均绝对误差百分比降低了6.7%,说明该方法在基坑变形预测方面具有一定的工程应用价值。

The foundation pit construction is the foundation and premise of many projects,and it is there⁃fore very important to predict its deformation characteristics scientifically.In this paper,a prediction method of unstable foundation pit based on EMD is proposed so as to understand the research status of neural network model in the deformation prediction of unstable foundation pit,further improve the predic⁃tion accuracy,and by referring to the idea of nonlinear and non-stationary signal processing.Firstly,the unstable deformation data is decomposed into sub-data sets with a certain change rules,and then the PSO-BP neural network model is used to predict the deformation of the sub-data sets.Finally,the pre⁃diction results of each sub-data set are superposed to obtain reliable prediction conclusions.The test re⁃sults show that the prediction accuracy of this method is significantly improved,the root mean square er⁃ror is only 0.417 mm,and the mean absolute percentage error is reduced by 6.7%,which shows that this method has certain engineering application value in the deformation prediction of foundation pit.