Effects of Missing Runway Performance Data on Maintenance Management Decision Making

Abstract

A problem commonly encountered in airport pavement management is how to deal with pavement condition and performance databases with various extents of missing data. While robust imputation techniques are now available to handle the problem of missing data, there are little to no guidelines available to assist airport engineers in evaluating the level of impact that data imputation has on the decisions made with respect to airport pavement maintenance. This research attempts to address the aforementioned issue by analyzing how different degrees of missing data would affect the maintenance decision concerning runway pavement sections in need of friction improvement. Using the runway friction data sample for illustration, 200 sets of data with varying degrees of missingness were generated. The generated data sets cover the following 10 levels of missingness: 5.55, 16.66, 25, 36.11, 44.44, 55.55, 66.66, 75, 86.11, and 91.66%. For each level of missingness, a total of 20 data sets with different patterns of missing data were randomly generated, and 20 plans of runway friction maintenance were derived with the help of imputed data. The derived maintenance plans, using data sets containing imputed data, were compared with the actual plan based on the original complete data set, and statistical hypothesis testing was performed to determine the reliability of maintenance plans derived using data sets containing missing data. The results of the analysis indicated a declining trend of reliability with increasing extent of missing data. Subsequent analysis of the results led to the conclusion that if the proportion of missing data remains under 40%, the derived maintenance plan is statistically indifferent from the actual plan at the 95% confidence level.