Multi-layer Gaussian Mixture Model algorithm and its application in lithology classification

doi:10.1080/21642583.2021.1961327

Return

Multi-layer Gaussian Mixture Model algorithm and its application in lithology classification

2021-08-04

DOI：10.1080/21642583.2021.1961327

Original

Save

View PDF

Abstract

En 中文

Even in the same block, the difference in lithology distribution is small, but there is still a big difference in the distribution of the central wells and the edge wells in the block at the same level of depth. This is because we need to find data points of the same stratum depth instead of being disturbed by data points of horizontal depths. According to this problem, we propose two GMM algorithms based on EM: multi-layer GMM and two-layer GMM. The feature of multi-layer GMM is that even if there are edge wells far away from the exploratory well, this algorithm can still filter out data points that are not in the same formation depth through continuous averaging of initialization information in order to achieve a better effect of accurately predicting the lithology distribution of the stratum. Moreover, the robustness is more suitable for the general situation. In addition, in the blocks with relatively concentrated wells, we also propose a two-layer GMM method complementary. Compared with the multi-layer structure, due to the parallel structure, it is faster but less robust. The simulation results show that the multi-layer GMM algorithm proposed in this paper is compared with the multi-resolution clustering method based on graph theory, single-layer GMM and double-layer GMM. This algorithm achieves a better clustering effect than other methods.

Keywords:

Log interpretation

lithology identification

Gaussian Mixture Model

clustering

log response characteristics

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.