Page 77 - Reliability and Maintainability of In service Pipelines

66 Reliability and Maintainability of In-Service Pipelines

This in-line inspection (ILI) technique magnetizes the pipe wall as it travels
down the structure and uses coil sensors to measure magnetic flux leakage inten-
sity along the pipe walls. MFL is based on an array of sensors placed along the
interior of the pipeline walls which detect changes or irregularities in the mag-
netic field. These irregularities are where potential defects may exist and can be
evaluated to determine what type of defect is present. The severity of the defects
is determined by estimated defect depth and consequently the safety of the pipe
by calculating maximum allowable operating pressure (MAOP) of oil or gas
through the pipes. Some common defects that this technique may identify include
corrosion, deformations, fatigue, hairline cracks, dents, buckles, delaminations,
and faulty welds.
Fig. 2.8 schematically illustrates how the sensors are placed on the pipe wall
and how the defect/crack can be detected. The tool includes the use of magnets
and steel couplers which touch the pipe wall to create a magnetic circuit. If the
wall thickness increases or decreases at a particular section, magnetic flux change
will be detected which allows the defect to be identified and later be evaluated
through the data acquired.
Since MFL technique can record changes in more than one or two directions,
it can detect even the small defects by most subtle changes of magnetic. After the
inspection is completed, using the MFL technique, data and image are transferred
to a computer monitor where the information is rendered and stored as 2D or 3D
images of the exterior or interior section of pipe walls.
Using the MFL technique, various submethods of pipeline inspection were
used for classification and regression of the infrastructure. The methods used
included Kernelization, regularized least-squares regression, support vector
machines for regression, kernel PCA, and partial least squares regression. The
goal of these procedures is to extract the most relevant features from a set of can-
didate features.

Figure 2.8 Performance mechanism of magnetic flux leakage technique.