Following the completion of in-line inspection (ILI) run comparison (RC), distributions of anomaly depth changes are interpreted to derive corrosion growth rate (CGR) estimates. However, these distributions inherently contain various sources of error. Addressing and accurately quantifying these errors presents a persistent challenge within the industry. This paper delves into an approach specifically designed to address one of the error sources linked to the RC algorithm's performance.

When faced with suboptimal RC performance, engineers may resort to a limited sample of manual matches that includes the deepest anomalies reported in both ILIs. This approach frequently results in overconservative CGRs, primarily because the sample tends to include anomalies that were oversized during the inspections. If suboptimal RC performance goes unnoticed, the resulting erroneous matches can significantly impede efforts to rectify inspection-related errors.

In this study, the performance specification of an experimental RC algorithm, based on the k-nearest neighbours principle, was utilized to mitigate the influence of matching errors on obtained CGRs. The performance of RC algorithm was established using a comprehensive dataset of synthesised anomaly matches. CGR statistics computed post-RC were contrasted with benchmark CGRs, pre-determined through various anomaly growth simulations. Discrepancies from this comparison informed the development of an error reduction strategy. Additionally, the effects of common change-in-depth filtering on rate estimations were examined, such as the exclusion of matched anomalies exhibiting negative delta depth.

The work presented demonstrates the potential to avoid overly conservative CGRs derived from two-inspection comparisons. Achieving more accurate corrosion growth rates paves the way for practical and better-prioritized repair strategies. However, the scope of this work is constrained by the limited number of simulations conducted. Before integrating the error-compensating approach into the complete CGR production process, more extensive testing is underway.