Predicting Lost Circulation in Drilling Using Machine Learning
Author Information
Author(s): Azadivash Ahmad
Primary Institution: Department of Petroleum Engineering, Amirkabir University of Technology, Tehran, Iran
Hypothesis
Can machine learning models effectively predict the intensity of lost circulation using well log data?
Conclusion
The study demonstrates that machine learning models can accurately predict lost circulation intensity, improving drilling efficiency and safety.
Supporting Evidence
- Machine learning models showed high predictive performance for lost circulation intensity.
- Random Forest and Extra Trees were identified as the best-performing methods.
- The study utilized a dataset of 1662 data points from well logs.
Takeaway
This study shows that we can use computers to help predict when drilling will lose mud, which can save time and money.
Methodology
The study applied seven machine learning methods, including Random Forest and Extra Trees, to predict lost circulation intensity based on well log data.
Potential Biases
The dataset had class imbalance, which could affect model performance.
Limitations
Some models showed limitations in predicting certain intensity classes, indicating a need for further refinement.
Participant Demographics
Data was collected from three wells in a gas field in northern Iran.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website