基于HGWO-SVR模型的竖向受荷斜坡桩基沉降预测

Settlement Prediction for Pile Foundation of Vertically Loaded Slope Based on HGWO⁃SVR Model

采用灰色关联分析深入研究了竖向荷载作用下斜坡桩基沉降的关键因素,各因素影响程度由大到小排序为:弹性模量>临坡距>斜坡坡度>内摩擦角>黏聚力>土体密度>土体泊松比>桩长>桩径。为优化支持向量回归(SVR)模型参数,引入差分进化,建立混合灰狼算法(HGWO),提出了一种新的HGWO-SVR模型。该模型与GWO-SVR和GS-SVR模型相比,表现出更显著的预测优势,整体预测精度高,误差较小。基于HGWO-SVR模型构建了斜坡桩基沉降的预测模型,并将其预测结果与已有沉降计算公式计算结果进行对比,结果表明,HGWO-SVR模型预测结果与公式计算结果最大误差为6.55%,验证了该模型在斜坡桩基沉降预测方面的可行性。

The key factors of pile foundation settlement were explored for the slope under vertical load by using grey relational analysis,and it is found that each factor is in the following descending order by its influence:elastic modulus>slope distance>slope gradient>internal friction angle>cohesion>soil density>poisson′s ratio of soil>pile length>pile diameter.In order to optimize the parameters of support vector regression(SVR)model,a novel HGWO⁃SVR model was proposed by integrating the differential evolution⁃enhanced gray wolf algorithm(HGWO).Compared with GWO⁃SVR and GS⁃SVR models,this model presents obvious advantage in prediction,with high accuracy and minor error.A settlement prediction model for pile foundation of slope was constructed based on HGWO⁃SVR model,and the prediction results were compared with those values calculated with existing settlement formulas.The results show that the maximum percentage error between the prediction value of HGWO⁃SVR model and the calculated value is 6.55%,thus verifying that this model is feasible in settlement prediction for pile foundation of slope.