中文摘要：

针对滑坡变形具有非平稳性、非线性与随机性变化特点,提出将小波分解与RBF神经网络相结合应用于滑坡变形预测。通过实验进行小波分解及不同低频-高频分量组合的预测,着重分析了不同的小波分解层数、分量组合形式以及预测步长对滑坡变形预测的影响。实验分析结果表明：只有选取适当的分解层数、合理的低频-高频分量组合与预测步长,才能得到最优的预测效果。同时也验证了本文方法的正确性,对于滑坡变形预测具有一定的参考意义。

英文摘要：

Based on the characteristics of non-stationary,non-linear and stochastic landslide deformation changes,a combination method of wavelet decomposition and RBF neural network is proposed for the landslide prediction. Based on experiments of wavelet decomposition and prediction of the combination of different low frequency and high frequency components,the effect of different wavelet decomposition levels,the component combination and predictive steps is analyzed. The experimental results show that only the appropriate decomposition level,proper component of combination and predictive step are selected,can we obtain optimal prediction. Also,the correctness of the method in the paper is verified. All these studies and results provide reference for the predication of landslide.