论文部分内容阅读
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。
Aiming at the model optimization problem of compaction curve, the AIC numerical regression method is proposed. The minimum AIC value of the integrated polynomial order and regression error is considered as the optimal compaction curve model. Through the analysis of 8 groups of experimental data, 4 groups are the best for the second-order polynomial fitting and the other 4 groups are the best for the fourth-order polynomial fitting. The results show that the compaction curves can not be described by the uniform order optimal model because of different compaction characteristics and test data points distribution. However, the model optimization method based on AIC criterion has good practicability.