The effect of visual inspection reliability on risk-based inspection

Bridge inspection has come a long way from its inception as a component level practice following the Silver Bridge collapse in 1967 to the most recent advancement enacted by the Moving Ahead for Progress in the 21st Century (MAP-21) legislation to determine the inspection interval using risk-based inspection (RBI) and mandated element-level data collection for all the states. RBI, a well-researched topic in other fields is new to the bridge inspection practice and element-level bridge inspection provides data for RBI condition attributes. The first objective of the research was to determine how inspection variability in condition state (CS) assignment and defect quantification affect the attributes in RBI practice. Based on the data collected from Indiana and Michigan, about 9 [percent] of the inspectors incorrectly assigned the bridges' elements to CS 4 (RBI screening attribute), and 18 [percent] of the inspectors incorrectly assigned the bridges' elements to CS 2 (low ranking score). The conclusion was made assuming that if most of the inspectors assign the same CS for a bridge's element, it is the correct CS assignment. The second objective was to determine how the inspection data variability affects the deterioration models (Markov chain, Kaplan-Meier, and Weibull). Based on the National Bridge Inventory (NBI) data from the seven states, it was shown that the NBI data do not fit the Weibull distribution and the variability in NBI data on Kaplan-Meier either extends or shortens the median time in condition rating (TICR) for a bridge component. Build on the details of the second objective, the Markov chain was proposed to calculate the probability of failure (POF) for the RBI occurrence factor (OF).