Performance of the synergetic wavelet transform and modified K-means clustering in lithology classification using nuclear log

Abstract

Accurate lithology identification is fundamentally crucial to reservoir evaluation from geophysical well logs. However, the traditional way of lithological identification is carried out in laboratory, which is not only expensive, but also time consuming in its interpretation. In this study, the synergetic wavelet transform and modified K-means clustering techniques are performed to classify metamorphic rocks from Chinese Continental Scientific Drilling Main Hole (CCSD-MH). At the beginning, different wavelet functions in different well logs are presented to detect lithologic interfaces. Meanwhile, the Haar wavelet and GR are determined to be the optimum wavelet function and well log, and the range of the optimum scales is about 8–15m in the reference well. After that, a fast and practical K-means clustering algorithm is employed to make a classification of stratigraphy into 5 groups, which are demarcated from the performance of wavelet transform. The results achieved are in accordance with the stratigraphic column and have a higher accuracy compared to the previous studies, indicating that the combination of the wavelet transform and modified K-means clustering can improve the accurate rate for the classification of metamorphic rocks in CCSD-MH.