去噪分析和分离变形预测在深埋隧道中的应用研究

隧道变形易诱发相应的工程问题,对其防治及预测已成为地下工程领域的热点问题。为提高隧道变形的预测精度,达到有效掌握隧道变形规律的目的,以某隧道为工程实例,通过对其监测数据的去噪处理,将隧道变形的原始序列分离为趋势项和误差项序列,并利用GA-BP神经网络和时间序列模型对两序列进行预测,又结合支持向量机模型对前者的预测误差进行修正,以保证预测精度。结果表明:在去噪方面,得出半参数优化卡尔曼滤波的去噪效果最优,其次是sym8小波去噪和奇异谱分析;在预测方面,得出分离预测能一定程度上提高预测精度,但效果不明显,而误差修正模型能很大程度上提高预测精度,综合得到本文预测结果的平均相对误差为1.08%。预测模型具有精度较高等优点,能为深埋隧道的变形预测提供借鉴。

Application of Denoising Analysis and Separation Deformation Prediction in Deep Buried Tunnel

The tunnel deformation tends to induce corresponding engineering problems, and the prevention and prediction of tunnel deformation has become a hot issue in the field of underground engineering. To improve prediction accuracy of tunnel deformation and effectively control tunnel deformation,the original series of tunnel deformation are separated into trend and error sequences with reference to a tunnel engineering based on monitoring data denoising. The prediction of the two sequences is conducted by using GA-BP neural network and time series model,and the prediction error of the former is corrected with the modified support vector machine model so as to ensure the accuracy. The results show that the denoising effect of optimal semi parameter optimization of Calman filter is the best,followed by Sym8 wavelet denoising and singular spectrum analysis; as far as prediction is concerned,the separation prediction can improve the prediction accuracy to a certain content,but the effect is not obvious; while,the error correction model can greatly improve prediction accuracy; and the average relative error of the prediction is 1. 08%. The prediction model has the advantage of high precision and can provide some reference for deformation prediction of deep buried tunnel.