Assessment of formation strength from geophysical well logs using neural networks

Jason E. Ressler, Christopher D.P. Baxter, Kathryn Moran, Meghan Paulson, Ion Ispas, Hans Vaziri

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


This paper presents the methodologies and results of two types of neural networks used to estimate the unconfined compressive strength (UCS) of weakly cemented sandstones from geophysical log data. The first neural network used 29 different logs as input and predicted UCS values at 8 cm depth resolution that were in good agreement with measured values. The second neural network used an innovative approach to improve the depth resolution and detection of thin bedding by relating changes in high resolution logs (e.g. bulk density or resistivity) to changes in strength. Preliminary analyses used to predict the undrained shear strength of marine clay illustrate the potential of this new approach for predicting the strength of weakly cemented sandstones. Copyright ASCE 2006.

Original languageEnglish
Title of host publicationGeoCongress 2006
Subtitle of host publicationGeotechnical Engineering in the Information Technology Age
Number of pages1
StatePublished - 2006
EventGeoCongress 2006 - Atlanta, GA, United States
Duration: Feb 26 2006Mar 1 2006

Publication series

NameGeoCongress 2006: Geotechnical Engineering in the Information Technology Age


ConferenceGeoCongress 2006
Country/TerritoryUnited States
CityAtlanta, GA


Dive into the research topics of 'Assessment of formation strength from geophysical well logs using neural networks'. Together they form a unique fingerprint.

Cite this