Direct Multi‐Modal Inversion of Geophysical Logs Using Deep Learning

Abstract

Geosteering of wells requires fast interpretation of geophysical logs which is a non‐unique inverse problem. Current work presents a proof‐of‐concept approach to multi‐modal probabilistic inversion of logs using a single evaluation of an artificial deep neural network (DNN). A mixture density DNN (MDN) is trained using the ”multiple‐trajectory‐prediction” loss functions, which avoids mode collapse typical for traditional MDNs, and allows multi‐modal prediction ahead of data. The proposed approach is verified on the real‐time stratigraphic inversion of gamma‐ray logs. The multi‐modal predictor outputs several likely inverse solutions/predictions, providing more accurate and realistic solutions compared to a deterministic regression using a DNN. For these likely stratigraphic curves, the model simultaneously predicts their probabilities, which are implicitly learned from the training geological data. The stratigraphy predictions and their probabilities obtained in milliseconds from the MDN can enable better real‐time decisions under geological uncertainties.

Real-Time Automated Geosteering Interpretation Combining Log Interpretation and 3D Horizon Tracking

“…The user then reviews and refines the generated horizon shapes and repeats this process for each horizon of interest. More recent versions of these algorithms leverage various comprehensive features within the 3D seismic volume, see [20,[22][23][24] and incorporate advanced data processing approaches like deep learning, see [25][26][27][28][29][30].…”

Machine Learning Developments and Applications in Solid‐Earth Geosciences: Fad or Future?

“…ML has also drastically improved the efficiency of geophysical signal processing and interpretation. Alyaev and Elsheikh (2022) use a mixture density deep neural network (NN) to perform fast geophysical log interpretation for real‐time geosteering. Automatic 2D image fault interpretations are performed on optical and topographic images with different resolutions (Mattéo et al., 2021), and bathymetry images (Vega‐Ramírez et al., 2021).…”