Estimation of elastic wave velocity and DCPI distributions using outlier analysis

Abstract

The reliability of a geostatistical method depends on the probability distribution of the measured values; unreasonable data should be removed. The objective of this study was to suggest a technique for predicting the distribution of geotechnical engineering data, namely the elastic wave velocity and the dynamic cone penetration index (DCPI), while eliminating biased values through outlier analysis. The refraction survey was applied to obtain elastic wave signals at four profiles, with lengths of 90 m, 20 m, 20 m, and 20 m, in the Geohwa Mountains, Korea. The Dynamic Cone Penetration Test (DCPT) was also performed at 15 sites to suggest a reference value of the objective area, and the measured values were converted into the DCPI. To determine the range of the biased value for the elastic wave and DCPI, the cross validation and the Generalized Extreme Value (GEV), that is, the Gumbel, Frechet, and Weibull functions, were selected, and confidence levels of 10%–90% were determined. A total of 8 and 2 data items were removed from the measured elastic wave velocity and DCPI after considering the confidence level through outlier analysis. Finally, contour maps of the elastic wave velocity and DCPI were created, and the enhancement was verified using the calculated performance ratio. The results suggest that outlier analysis based on cross validation and the GEV can provide highly reliable distributions of geotechnical engineering data with implementing geostatistical methods.