Real-time prediction of Poisson's ratio from drilling parameters using machine learning tools

Area of Science:

• Geophysics
• Petroleum Engineering
• Machine Learning

Background:

• Poisson's ratio is crucial for wellbore stability, stress estimation, drilling performance, and hydraulic fracturing.
• Traditional methods for Poisson's ratio estimation involve costly and time-consuming laboratory experiments or sonic logs.
• Limited availability of data and samples are significant drawbacks of conventional techniques.

Purpose of the Study:

• To present a novel, real-time technique for predicting Poisson's ratio using readily available drilling parameters.
• To evaluate the effectiveness of machine learning algorithms, specifically Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS), for this prediction.
• To demonstrate a cost-efficient alternative to conventional methods for geomechanical property estimation.

Main Methods:

• Utilized real-time drilling parameters: weight on bit, penetration rate, pump rate, standpipe pressure, and torque.
• Employed Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) machine learning algorithms.
• Trained and validated models using extensive datasets (2905 and 2912 data points) from two distinct wells with diverse lithologies (carbonate, sandstone, shale).

Main Results:

• Achieved high correlation coefficients for predicted Poisson's ratio: 0.98-0.99 for ANN and 0.97-0.98 for ANFIS.
• Reported low average absolute percentage errors: 1-2% for ANN and approximately 2% for ANFIS.
• Demonstrated a strong agreement between actual and predicted Poisson's ratio values.

Conclusions:

• Drilling data combined with machine learning offers a powerful tool for real-time geomechanical property prediction.
• The proposed method significantly reduces costs and time compared to conventional laboratory or sonic log approaches.
• This technique enables efficient, in-situ estimation of Poisson's ratio during drilling operations.