Spatial Statistical “Blind-Approach” Results

This chapter presents the results from the statistical “blind-approach” proposed in Chapter 6 . As mentioned above, three main scales were considered: a full, segmented, and defect-based scale. These scales do not intend to make multiscale modeling, i.e., to assess the corrosion degradation in a coupled way; on the contrary, the intention is to analyze different scales of the condition of the case study. The full-scale analysis implements the seven soil categories mentioned before to detect which conditions tend to be more aggressive at the outer wall. The analysis was also implemented with the data at the inner wall for comparison purposes. In the inner wall, the influence of the soil parameters would be diminished, and no influence is expected. The segmented scale uses the changepoint approach to divide the pipeline based on the depth, length, and width measurements. The segmentation allows decision-makers to focus on critical segments instead of the entire pipeline. Finally, the defects scale uses Moran’s scatterplot and correlogram and the relationship between the areas and volumes of the neighbors of each defect to inspect their spatial autocorrelation. This autocorrelation allows identifying further clusters where spatial independence assumptions are no longer valid, and further spatial predictions can be proposed.