Integrating multiple seismic attributes for fault detection using a new hybrid machine learning.

Area of Science:

• Geophysics
• Machine Learning
• Data Science

Background:

• Seismic data analysis is crucial for subsurface exploration.
• Accurate fault detection in seismic data presents challenges due to complex patterns and noise.
• Existing methods may have limitations in recognition accuracy and error reduction.

Purpose of the Study:

• To develop a novel hybrid machine learning approach for enhanced fault detection in seismic data.
• To improve the recognition of fault patterns and reduce detection errors.
• To combine the strengths of Multilayer Perceptron (MLP) neural networks and Support Vector Machines (SVM) for superior performance.

Main Methods:

• Integration of Multilayer Perceptron (MLP) neural networks with Support Vector Machines (SVM) for a hybrid fault detection model.
• Extraction and normalization of diverse seismic features including GLCM, ant tracking, chaos, variance, sweetness, correlation, slope direction, and energy.
• Training and evaluation of independent MLP and SVM models, followed by combining their predictions.

Main Results:

• The hybrid approach demonstrated significantly increased fault detection accuracy compared to individual MLP and SVM models.
• The combined predictions from MLP and SVM models led to improved overall accuracy.
• The study successfully enhanced the analysis and recognition of fault patterns in both synthetic and real seismic data.

Conclusions:

• The proposed hybrid machine learning method offers a robust solution for accurate seismic fault detection.
• Integrating MLP and SVM models effectively leverages their respective strengths for improved pattern recognition.
• This approach provides a valuable tool for geoscientific exploration, reducing detection errors and enhancing data interpretation.