基于变权重组合预测模型的软土路基沉降预测研究

Study on the Settlement Prediction of Soft Soil Subgrade Based on Variable-weight Combination Prediction Model

为建立更加符合软土路基沉降的预测模型与方法,以某公路工程实测沉降数据为研究对象,对软土路基沉降规律进行分析基础上,分别建立Logistic模型,双曲线模型,Gompertz模型,乘幂模型和对数模型5种预测模型.将5种单项模型预测值与实测值进行对比分析,得到3种优势模型.以组合预测模型理论为依据,赋予3种优势模型不同权重系数建立变权重组合预测模型.结果表明,模型预测值与实测值拟合程度更好,预测精度和可靠性都显著高于单项预测模型,具有明显的优越性.模型能准确地预测长期沉降,更适合于作为公路软土路基沉降预测模型.

In order to establish a more suitable prediction model for soft soil subgrade settlement,the actually settlement measured data of a highway project was taken as the research objective.According to the regularity shown of soft soil subgrade settlement,five prediction models were used to fit the measured settlement data,including Logistic model,hyperbola model,Gompertz model,power model and logarithmic model.Comparison and analysis of the predicted values and the measured values about the five prediction models,three dominant models were obtained.Based on the theory of combinatorial forecasting model,the variable-weight combinatorial forecasting model was established by assigning different weight coefficients to three dominant models.The results show that the predicted value obtained by this model can fit well the measured value.Both accuracy and reliability were significantly higher than single prediction models,this model can accurately predict the longterm settlement,which is more suitable for the use of the highway soft soil subgrade settlement prediction.