Identification and Modeling of New Defects

The statistical “blind-approach” in the last chapter focused on how spatial variability could be described at three main scales (i.e., full, segment, and defect-based). This approach considered the data reported from at least one ILI measurement and basic soil classification. However, new defects will appear between consecutive inspections, mainly because of the aggressiveness of the surrounding soil, the fluid being transported, and the inspection capabilities. Dealing with new defects implies two main processes determining the number of defects for a given time and spatial distribution. Based on the above, this chapter proposes an approach to identify and characterize new defects between two consecutive inspections. The characterization considers how they are distributed, interact with old defects, and how new defects can be simulated. For this purpose, a point pattern analysis is proposed based on the reported locations of each corrosion defect. This chapter considers a matching process between the two inspections to identify the defects not reported in the first inspection. This process is not straightforward because defects may be shifted from their real location (i.e., location errors). Hence, point pattern matching seeks a feasible transformation that maps the location of the defects from the first to the second inspections. Besides, defect distribution is evaluated considering the Complete Spatial Randomness (CSR) assumption, and the new defects are spatially predicted based on a non-homogeneous Poisson process.