Use of artificial intelligence tools to identify CaCO3 polymorphs in the scaling process under oil well conditions

Abstract

Inorganic scaling by calcium carbonate (CaCO₃) in oil wells is one of the major challenges to flow assurance, especially in pre-salt environments where variations in pressure, temperature, and CO₂ degassing promote the precipitation of polymorphs such as calcite, aragonite, and vaterite. These deposits reduce production efficiency and significantly increase operational costs. In this context, this study proposes the use of artificial neural networks (ANNs) and computer vision techniques to automatically identify and segment CaCO₃ polymorphs from in situ microscopic images. The model was developed using the YOLOv8n-seg architecture, a state-of-the-art You Only Look Once (YOLO) convolutional neural network designed for real-time object detection and segmentation with high accuracy and low computational cost, adapted here to identify and quantify CaCO₃ crystalline morphologies. The training, performed with images obtained in a pressurized reactor under simulated pre-salt well conditions, achieved expressive performance metrics (mAP@0.5 = 0.933 and F1-score = 0.89). The results demonstrate that the Artificial Intelligence (AI) based approach can distinguish calcite, aragonite, and vaterite morphologies quickly and reliably, complementing traditional characterization methods such as X-ray diffraction and scanning electron microscopy, by enhancing automatic and real-time analytical capabilities. The study highlights the potential of neural networks to improve the diagnosis and control of scale formation in petroleum production systems, contributing to higher operational efficiency and safety.