基于GM-RBF组合模型的高路堑边坡稳定性预测研究

Study on stability prediction of high cutting slope based on GM-RBF combination model

边坡的稳定性预测是道路边坡安全性评判的关键,且及时、准确的预测可以有效地预防边坡破坏灾害的发生。采用GM(1,N)模型与RBF神经网络模型相结合的方式,建立一种基于GM-RBF组合的高路堑边坡变形预测分析模型。结合高速公路高路堑边坡工程实例,通过对比GM(1,5)模型、RBF神经网络模型和GM-RBF组合模型的边坡安全系数预测结果来分析GM-RBF组合模型的可行性。结果表明,GM-RBF组合模型比单一模型更能抵抗预测数据序列中存在的波动性;较于GM(1,5)模型和RBF神经网络模型,GM-RBF组合模型预测的边坡安全系数平均绝对误差分别降低了64.6%和45.8%,边坡安全系数均方根误差分别降低了66.7%和45.2%,边坡安全系数相对均方误差也分别降低了58.3%和38.7%;采用GM-RBF组合模型对边坡稳定性进行预测能够保持良好的精度。

Prediction of slope stability is the critical point of the road slope safety evaluation, and timely and accurate prediction can effectively prevent the occurrence of slope damage disasters. The analysis model of deformation prediction for high cutting slope based on GM-RBF combination mode was established by adopting the combination of GM(1,N) model and RBF neural network model. Combining the high cutting slope engineering example of the expressway, the feasibility of GM-RBF combination mode was analyzed by comparing the prediction results of the slope safety factor of GM(1,5) model, RBF neural network model and GM-RBF combination model. The results show that GM-RBF combination model can resist the volatility in the predictive data series more effectively than single model. Comparing to GM(1,5) model and RBF neural network model, the average absolute error of slope safety factor predicted by GM-RBF combination model is reduced by 64.6% and 45.8% respectively, the root mean square error of slope safety factor is reduced by 66.7% and 45.2% respectively, and the relative mean square error of slope safety factor is reduced by 58.3% and 38.7% respectively. Using GM-RBF combination model to predict slope stability can maintain good accuracy.