基于多模型加权组合算法的公路边坡位移预测

Prediction of Highway Slope Displacement Based on a Weighted Multi-model Combining Algorithm

基于边坡位移实际监测数据序列对未来某一时间的位移进行预测是一种有效且实用的边坡稳定性分析方法,将不同的预测方法加以组合,可以提高预测结果的可靠度。对此,分析了常用边坡位移预测方法的基础之上,提出一种多模型加权组合预测算法。先分别用不同的预测模型对边坡位移进行预测,然后根据预测结果的方差以一定的方法对每种预测模型的预测结果加权,最后求和得出最终位移预测值。算例表明,预测方法预测结果的相对误差均值要小于单一预测模型,和同类文献中的算法对比,在保证了预测精度的同时亦能大大提高算法效率。

To predict the future time slope displacement based on the displacement actual monitoring data sequence is an effective and practical method for slope stability analysis. And combining different prediction methods can improve the reliability of prediction results. Therefore,this paper has analyzed the current displacement prediction model,then proposed a weighed multi-model prediction algorithm. First,using different prediction models to predict the slope displacement. Then,assigned a weight for each prediction model with a certain method according to the variance of prediction results. Finally,weighted and summed the prediction results then the final displacement predictive value can be drawn. The numerical examples showed that this method is superior to a single model. And compared to similar algorithms in the literature,the accuracy can be ensured. However,the algorithm efficiency has been greatly improved.