Use of artificial intelligence to identify CaCO3 polymorphs during the scaling process in oil wells.
Nombre: OZEAS DOS SANTOS SILVA SOUZA
Fecha de publicación: 27/07/2026
Junta de examinadores:
Nombre |
Rol |
|---|---|
| DANIEL DA CUNHA RIBEIRO | Examinador Interno |
| FABIO DE ASSIS RESSEL PEREIRA | Presidente |
| GISELLE MARIA LOPES LEITE DA SILVA | Examinador Externo |
| WANDERLEY CARDOSO CELESTE | Coorientador |
Resumen: Calcium carbonate (CaCO3) scaling in oil wells, composed of the polymorphs calcite, aragonite, and vaterite, can compromise the effectiveness of completion over time and significantly reduce the productive capacity of wells. This study proposes the use of artificial intelligence, applying computer vision through the YOLO architecture for the automated instance segmentation of
these polymorphs in optical microscopy images, an approach unprecedented in the consulted literature for the flow assurance domain. The accurate identification of these crystalline forms is essential, since their different physicochemical properties influence the scaling process under different thermodynamic conditions. Therefore, the YOLO model, trained with images of CaCO3 samples, provides an automated and reproducible approach for the analysis of these deposits. The results demonstrate that the model achieves an mAP@0.5(M) of 96.9% and an aggregate F1 of 0.94 on instance segmentation, with zero inter-class confusion in polymorph classifica- tion. In addition, a post-processing pipeline converts the segmentation masks into calibrated
morphometric descriptors, of which the DF/Deq ratio constitutes the primary discriminator between polymorphs: aragonite presents a mean ratio of 1.99±0.38, consistent with acicular habit, and calcite, 1.33±0.26, consistent with rhombohedral arrangement, with a separation corresponding to 2.54 standard deviations between distributions. Thus, the study demonstrates
the potential of computer vision combined with quantitative morphometry to provide automated and reproducible diagnostics oriented to flow assurance, contributing to the standardization of inorganic scale analysis in the oil and gas sector.
