Eventos Anais de eventos
CONEM 2022
XI Congresso Nacional de Engenharia Mecânica - CONEM 2022
Automatic Identification of Flanged Joints in Pipeline Systems in the Oil and Gas Industry
Submission Author:
Nathalia Nascimento , RJ
Co-Authors:
Nathalia Nascimento, Alessandro Zachi, Milena Pinto
Presenter: Alessandro Zachi
doi://10.26678/ABCM.CONEM2022.CON22-0189
Abstract
In recent years, the offshore Oil & Gas industry has increased the search for automated processes to ensure efficiency and quality in manufacturing. With new robotic technologies, traditional operations were moved towards more autonomous and advanced processes, reducing operational costs and human effort, increasing safety and reliability. It is well-known from the literature that regular underwater inspections can sometimes be difficult to accomplish, very expensive to maintain, and also be time-consuming. Moreover, such kind of inspection is generally carried out by using Remotely Operated Vehicle (ROVs) equipped with end-effector tools, spotlights, sensors and appropriate cameras. However, in such a general inspection approach, the human operator must guide the vehicle safely through the structure while performing the visual inspection toward detecting possible structural faults and movements if necessary. In such a context, computer vision technology provides a suitable non-contact approach to the real-time problem of detection and has emerged as a potential tool in robotics sensing field. Therefore, this work aims to detect flanged joints from image scenes collected during the inspection services of piping systems that can be both operated onshore and offshore. The work also presents and discusses a computational intelligence algorithm specifically developed for extracting the data from the captured images and for providing decision making. By characterizing the presence of a connection region in the underwater structure, the purpose of searching for flanged joints is to streamline the work of the operator who would previously perform a manual search. Some preliminary tests have been carried out by using a given image dataset for feeding the proposed algorithm. The preliminary results obtained showed that the use of the proposed machine learning-based technique has achieved satisfactory metrics that show certain interesting and attractive efficiency levels that would characterize it as a potential tool for carrying out this type of underwater inspection task.
Keywords
Subsea Mechanical Systems, Computer Vision, offshore industry, Inspection, Convolutional neural networks
