基于松散型小波神经网络的某隧道仰坡变形趋势研究  

Research on Deformation Trend of Tunnel Upward Slope Based on Loose-type Wavelet Neural Network

作　　者：隆然[1] 庞建成 

机构地区：[1]中交第二公路勘察设计研究院有限公司,武汉430056 [2]武汉华中岩土工程有限责任公司,武汉430023

出　　处：《路基工程》2016年第3期25-30,共6页Subgrade Engineering

基　　金：交通运输部西部交通科技项目(2011318493720)

摘　　要：针对某隧道洞口仰坡的稳定性问题,首先选取松散型小波神经网络对仰坡的沉降变形进行预测,再利用秩相关系数检验和Mann-Kendall检验对仰坡的变形趋势进行判断;结果表明：不同小波函数的去噪效果不一致,其中监测点～#5采用db 4小波的去噪效果最好,而监测点～#6则是采用db7小波的去噪效果最好;另外,秩相关系数检验和Mann-Kendall检验在仰坡变形趋势的判断中具有很好的一致性,均得出两监测点的变形趋势是往减小方向发展,仰坡变形趋于稳定,可以安全地进行后期施工。Aiming at the stability of upward slope at the tunnel entrance,the loose-type wavelet neural network is selected to predict the settlement deformation; the rank correlation coefficient test and MannKendall test are performed to judge the deformation trend of upward slope. The results show that the denoising effect of different wavelet functions is inconsistent; wherein,the denoising effect of db4 wavelet at monitoring points～#5 is the best,and the denoising effect of db7 wavelet at the monitoring point～#6 is the best. In addition,the rank correlation coefficient test and the Mann-Kendall test have very good consistency in the judgment of the deformation trend of upward slope. The deformation trend of two monitoring points is gradually decreased;and the deformation of upward slope is stable,so later construction can be done in a safe manner.

关 键 词：仰坡 BP神经网络 秩相关系数 小波去噪 MANN-KENDALL检验 

分 类 号：TU433[建筑科学—岩土工程]