最优组合预测模型在高填方体沉降中的应用研究

Research on application of optimal combination prediction model in settlement of high fill body

针对Logistic和Gompertz路堤沉降预测模型对于高填方土体后期沉降量预测精度较差的问题,该文基于Logistic和Gompertz预测模型的共性和个性,在充分考虑二者各自优势和特定局限性的基础上,对二者进行了最优化组合。以组合模型的最小对数误差平方和为目标函数求解最优加权系数,进而推导出最优加权几何平均组合预测模型,以提高高填方土体沉降预测精度的置信度。通过各预测模型对实际高填方土体工程的实测沉降数据进行拟合分析,验证最优组合预测模型在高填方土体沉降预测中的可行性。结果表明:最优组合预测模型的精度及可靠性均优于任一单一模型,适应性更强,尤其当监测时间超过曲线拐点后,几乎接近实测值。组合预测模型能够适用于多个模型优化组合,可涵盖各单一模型的优劣势,只要在权重比不出现负值的情况下,可大幅度的提高沉降预测精度。因而此组合预测模型可作为高填方土体后期沉降预测的一种有效方法。

Aiming at the problem of poor precision of two common settlement prediction of Logistic model and Gompertz model for predicting the final settlement of high fill body.Based on the commonness and individuality of Logistic model and Gompertz model,in order to give full play to their respective prediction advantages and complement their specific limitations,this paper optimizes the combination of the two models.Solving the optimal weighting coefficient with the objective function of minimizing the sum of squares of logarithmic errors of the combined model,and the optimal weighted geometric mean combination forecasting model is derived,so as to improve the confidence degree of the prediction accuracy of the settlement of the high fill..The feasibility of the optimal combination forecasting model in engineering prediction is verified by fitting the monitoring data of specific high fill projects with each settlement prediction model.The results show that the precision and reliability of the optimal combination prediction model are better than any single model,and the adaptability is stronger.Especially,when the monitoring time exceeds the curve inflection point,it is almost close to the measured value.The combined prediction model can be applied to the optimization combination of multiple models,and can cover the advantages and disadvantages of each single model.As long as the weight ratio does not have negative value,the accuracy of settlement prediction can be greatly improved.so the combined forecasting model can be used as an effective method for predicting the settlement of high fill embankment in the later stage of engineering.