Answers are grounded in this paper and cite their sources.
Depth alignment as a prerequisite for subsurface geology interpretation aims to match signals of two logs from the same well to a common depth scale. Here, three consecutive unsupervised machine-learning techniques are applied to achieve this goal. First, data vectors are reduced through pre-processing with PCA and autoencoder methods, and clustered by K-means with K-means++ initialization. In a second step, corresponding cluster boundaries are identified to globally align lithologic units. This assumes that cluster boundaries represent physical changes in lithology, without direct lithological interpretation. The third step uses cross-correlation to obtain the final log alignment. Additional linear or non-linear interpolation between matching boundaries is sufficient to align well logs. Performance and accuracy of various techniques are compared by visual inspection and manually determined boundaries. Reducing data dimension with an autoencoder and K-means clustering performed best among tested methods. The study suggests that cluster-based methods can automate depth alignment of well logs without human intervention.
No takes yet. Share an insight, caveat, or question.
Acharya et al. (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: