基于AIC准则的击实试验曲线拟合模型优选

Fitting model optimization of the compaction test curve based on the AIC criterion

针对击实曲线的模型优选问题,提出了基于赤池信息量准则(AIC)数字回归方法,认为综合多项式阶次与回归误差的AIC最小的模型为最优击实曲线模型。通过8组试验数据分析,4组为二次多项式拟合最优,而另4组以四次多项式拟合为最优。优选结果说明,由于不同土体压实性能和试验数据点分布不同,击实曲线不可能用统一阶次的最优模型描述,而基于AIC准则的模型优选方法具有较好的实用性。

In view of the problem of compaction curve model optimization,the study applied the AIC criteria polynomial regression algorithm to determine the optimal fitting model.The result shows that the best compaction curve fitting model is the one with the smallest AIC value,which combines the polynomial order and the regression error into consideration.Through 8 sets of test data analysis,4 groups are the best for the two-polynomial model,while the other 4 groups are the best for the four-polynomial model.Because the compaction performance of different soil is different from the point distribution of test data,it is impossible to describe the compaction curve with the uniform order,but the model selection method based on AIC criterion has good practicability.