路基压实质量连续检测数据前处理及修正方法

以兰张铁路工程为例，提出异常值的剔除方法，并修正压路机行驶速度与振动方式对连续检测数据(CMV)的影响。在分析现场异常数据产生原因的基础上，采取绝对中位差法、孤立森林法、局部异常因子识别法对异常值进行分析，并结合动态变形模量(E_vd)对筛选后的数据进一步分析。结果表明：CMV检测原理及压路机工作参数可作为识别处理缺失值和伪异常值的依据，绝对中位差法、孤立森林法对突变数值的处理效果最好。采用修正系数对CMV进行修正后，表明当压路机以1.2 m/s为行驶速度基准值时，可分别建立强振和弱振作用下的CMV值修正公式以实现提高检测可靠性的效果。

Pre-processing and Correction Method of Continuous Detection Data for Subgrade Compaction Quality

Taking the Lanzhang Railway Project as an example, a method for excluding outliers is proposed, and the influence of speed and vibration mode of rollers on the continuous detection data is corrected. Based on the analysis of the causes of outliers, the Median Absolute Deviation, Isolated Forest and Local Outlier Factor were taken to analyze outliers. Combined with dynamic deformation modulus (E_vd), the filtered data were further analyzed. The research shows that the principle of (CMV) and the roller parameters can be used as evidence for identifying and dealing with the missing values and pseudo-outliers, and the Median Absolute Deviation and Isolated Forest are the best methods for dealing with mutation values in outliers. After correcting the（CMV) with the correction factor, it is shown when the rollers takes 1.2 m/s as reference speed, the CMV correction formulas can be established respectively under strong and weak vibration modes, which can achieve the effect of improving the detection reliability.