Detection and Dispersion Analysis of Water Globules in Oil Samples Using Artificial Intelligence Algorithms

Area of Science:

• Petroleum Engineering
• Computer Vision
• Materials Science

Background:

• Accurate fluid particle detection is crucial for oil and gas industry operations, impacting refining techniques and equipment quality assessment.
• Existing methods for analyzing water globules in oil can be labor-intensive and may lack precision.

Purpose of the Study:

• To develop and validate a computer vision algorithm for detecting and analyzing water globules in oil samples.
• To assess the performance of a Convolutional Neural Network (CNN) based model for fluid particle analysis.

Main Methods:

• Development of a custom dataset of microphotographs from oil refinery samples.
• Implementation of the YOLOv4 Convolutional Neural Network (CNN) algorithm for object detection.
• Application of an authors' augmentation algorithm to increase the training dataset size.
• Dispersion analysis and frequency diagram generation for particle size distribution.

Main Results:

• The developed YOLOv4-based algorithm accurately detects water globules in oil samples.
• The model achieved high accuracy metrics: AP@50 = 89% and AP@75 = 78%.
• Dispersion analysis provided insights into the size and number distribution of detected water particles.

Conclusions:

• The CNN-based model is verified and suitable for detecting particles in fluid media.
• The algorithm offers a reliable and accurate method for evaluating fluid quality and refining processes.
• The developed tool provides researchers with controllable accuracy for particle detection.