Automatic Implementation Algorithm of Pressure Relief Drilling Depth Based on an Innovative Monitoring-While-Drilling...

Area of Science:

• Geotechnical Engineering
• Signal Processing
• Machine Learning

Background:

• Monitoring-while-drilling (MWD) methods are crucial for efficient drilling operations.
• Identifying the amplitude concentrated enlargement zone in vibration signals indicates drilling depth.
• Existing methods lack high accuracy in automatically identifying this zone.

Purpose of the Study:

• To develop a high-accuracy model for automatic identification of the amplitude concentrated enlargement zone.
• To improve the accuracy of determining drilling depth from vibration signals.
• To provide a foundation for automated drilling depth identification.

Main Methods:

• Proposed a novel neural network model combining Deep Neural Network (DNN) and Long Short-Term Memory (LSTM) networks.
• Utilized both single-sensor and multi-sensor data for prediction.
• Implemented an optimization method to mitigate data anomalies.

Main Results:

• Single-sensor identification accuracy reached 92.72%.
• The proposed multi-sensor neural network model achieved accuracy greater than 97.00%.
• Optimization improved accuracy to the level of manual recognition.

Conclusions:

• The developed neural network model effectively identifies the amplitude concentrated enlargement zone.
• This research solves a key challenge in automated drilling depth identification.
• The findings support the advancement of MWD technologies.