Abstract
Flexible pipelines are widely used for oil and gas production in deep waters across the national territory. They have been designed to withstand a great amount of stress due to internal and external pressure, wave forces and their own weight. Over time, these pipelines may present failures within their layers responsible for sustaining axial forces, known as tensile armor. Failures of this type can have serious economic and environmental consequences, in addition to high repair costs and, consequently, production losses. The adopted tensile armor is composed of metallic wires arranged in a helix format. They may break more frequently at the ends of pipelines, especially at the connection point of the riser with the floating production unit, where the stress concentration is higher. Initially, the breakage of these wires causes a twist in the pipeline and a deformation in its outer layer. As more wires break, the pipeline eventually loses its ability to operate and may collapse catastrophically.
The goal of this work is to detect and predict possible deformations and twists on risers already installed in floating production units, without replacing equipment or incurring any downtime. We propose a novel computer vision solution for visual monitoring of flexible pipelines that works by attaching two panels with planar patterns on the riser, and visually observing them over time. To detect deformation, an estimation algorithm was developed to be used together with a state-of-the-art technique for the detection of planar patterns.
Laboratory tests demonstrate that it achieves adequate precision in the level of deformation and that its algorithms are sufficiently robust to withstand the effects of light distortion in the calculation of deformations. Regarding the testing of the impact of camera positioning, the results showed that rotating around the axis of a pipeline is equivalent to using both rotation and pitch. This simplifies the capture procedure and is beneficial for the operation, in view of the difficulty in manipulating the pitch of the ROV and the camera while ensuring that the patterns remain within the field of view. As part of the accuracy evaluation, all likely sources of errors were considered as well as the impact of each of these regarding the accuracy of the established metrics. In conclusion, with a 99% confidence level, the proposed solution achieves an accuracy of 0.059 degrees per meter when estimating torsion.
This work provides a novel solution for measuring riser torsion, that can be applied to risers already installed, without incurring any downtime, while it also provides high precision results. The solution works both on aerial and underwater parts of the riser, as long as the capturing device can move along the medium. The main limitation of this solution is that it monitors only the region of the pipe where the panels with the planar pattern are attached to. It is seen as a comparative measurement solution, one that measures deformation relative to the first instance of data acquisition. Another important characteristic is that it is not a continuous monitoring system, as it requires executing a full video capture in order to compute the deformation.
