基于速率倒数法的滑坡时间概率预测

Probabilistic Forecast of Slope Failure Time Based on Inverse-velocity Method

在滑坡时间预测中,基于变形监测数据的速率倒数法受到了广泛重视。在该方法的使用过程中,存在模型参数难以同时标定、难以考虑预测时间的不确定性的难题。针对上述不足,提出了一种两阶段速率倒数模型的标定方法。基于该方法可以对模型参数进行同时标定,同时可对滑坡时间进行概率预测。结合10个滑坡案例比较了不同模型假设对滑坡时间预测的影响。结果表明,考虑滑坡时间的不确定性后,预测滑坡时间与实际观测滑坡时间更为符合。考虑速率倒数模型的非线性后,拟合最优模型与实测数据的拟合度更好;但从预测效果而言,线性模型预测结果与滑坡实际发生时间符合的程度更高。

The monitoring data-based inverse-velocity method has drawn wide attention in slope failure forecast.There are difficulties in calibrating model parameters simultaneously and considering the uncertainty of slope failure time.To solve such problem,a two-step method for calibrating the parameters of the inverse-velocity method is introduced.Meanwhile,the probabilistic prediction is provided.Based on 10 cases of landslide,the impact of the selection ofαvalue in the inverse-velocity method is discussed.The result shows that:(1)the predicted failure time can well accord with the actual time by considering its uncertainty with probabilistic method;(2)the linearized model can provide a more accurate forecast though the non-linearized model shows higher goodness-of-fit.