Outlier Analysis for Correlation Verifying Test of Roller Integrated Compaction Measurement

Identifying the outliers from the correlation verifying test data is essential to the roller-integrated compaction measurement (IRICM). Due to the variant filler properties, lift thickness, moisture ratio, and various roller modes, outliers may arise and result in the misunderstanding of the relationships between the roller-integrated compaction measurement values (IRCMV) and in-situ compaction test values (IICTV). The distance-based outlier model was established to characterize the dispersion degree of the data. Then the estimation neighbor method was used to identify the out-of-trend points, the outliers. The RICM field experiment was conducted in the calibration strips to test and verify the proposed method. The results revealed that the proposed method enabled the correlation and regression coefficients more accurate. This research demonstrates that the outlier analysis techniques for data preprocessing could be applied to RICM.