Analysis of Reliability and Effectiveness of Repeated Inspections Based on Probability of Detection Method

Abstract

Structural health monitoring (SHM) performs the function of evaluating performance and durability for structural life management by monitoring changes in engineering structures such as buildings and bridges. In order to obtain information about a structure's ability to perform its intended function, data collection activities are required through various inspections aimed at detecting the presence of structural damage. Repeated inspections have been proposed to increase the reliability of SHM. Many people considered repeated inspection as a way to increase the chance of detecting damage. If more than one of the individual inspections finds damage, collectively evaluates the damage as detected and produces the results of repeated inspections. Probability of detection (POD) was used as a measure of the sensitivity and reliability of the inspection process. To evaluate structural condition and predict remaining service life, POD is measured and structural life is calculated based on initial defect sizes that are just below the inspection limits of non-destructive testing techniques. Repeated inspections can be considered by multiplying the likelihood function, but if a single inspector performs repeated inspections, they may not be independent because they may be biased by previous inspection results. The repeated inspections is independent if performed by an automated system or another inspector unaware of previous inspection results. It can be assumed that each inspection is independent in that the SHM system can automatically collect data even in areas where general non-destructive testing is impractical due to complex geometries and accessibility limitations, but conversely, due to the dependencies of the data, there is no statistical difference between subsequent measurements. It is also considered to be less independent. In this paper, the effect of repeated inspection on POD improvement was confirmed using eddy current inspection data, and the benefits of repeated inspection differed from those predicted by assuming complete independence. Furthermore, the effectiveness of repeated inspection was discussed.