应用不同去噪方法及分形理论判断滑坡变形趋势

Application of different denoising methods and fractal theory in landslide deformation tendency prediction

为达到准确判断滑坡变形趋势的目的,首先采用多种去噪方法对滑坡变形数据进行去噪处理,分离滑坡变形的趋势项和误差项数据,并采用分形理论对各序列的变形趋势进行判断和对比研究;然后,再利用神经网络对滑坡变形进行预测以验证滑坡变形趋势判断的准确性。结果表明:不同去噪方法的去噪效果具有较大的差异,其中半参数卡尔曼滤波的去噪效果最优,且滑坡具有变形加剧,稳定性减弱的趋势;同时,对比趋势判断变形预测结果,两者的一致性较好,证明了滑坡变形趋势判断的准确性。研究成果可为滑坡的变形趋势判断提供一种新思路。

In order to predict landslide deformation tendency accurately, firstly, we used a variety of denoising methods to denoise the landslide deformation data and separate the landslide deformation tendency data and error data, and took fractal theory to judge and compare the deformation tendency of each sequence. Secondly, in order to verify the judgment accuracy of the landslide deformation tendency, we used neural network to predict the landslide deformation. The results showed that the denolsing effects by different denoising methods were quite different, the denoising effect of semi parametric Caiman filtering was the most optimal, and the landslide has a tendency of greater deformation and worse stability; meanwhile, compared with the tendency judgment of deformation prediction results, the consistency was good, proving the accuracy of the landslide deformation tendency judgment. This study could provide a new idea for judging the landslide deformation tendency.