AUGMENTED STOCHASTIC MULTIPLE IMPUTATION MODEL FOR AIRPORT PAVEMENT MISSING DATA IMPUTATION

摘要

This paper presents a research study to handle the problem of missing data in airport pavement management systems. This study is a continuation of an earlier study addressing the same concern. In the earlier study, a Stochastic Multiple Imputation (SMI) approach was adopted to overcome major limitations associated with conventional data imputation methods. The SMI approach considers the variation of multiple plausible imputations and obtains an unbiased estimate in replacing a missing data value. This approach was found to out-perform the three most commonly used imputation methods for missing data handling in pavement management: linear interpolation method, substitution by mean method, and regression method. However, the SMI approach estimates missing data values purely based on statistical techniques, without making use of any unique characteristics of pavement performance data. The present study explores the possibility to further improve the data imputation process by exploiting parallel pavement performance related data available from pavement condition and performance surveys of airfield pavements. An augmented stochastic Multiple Imputation (ASMI) approach is proposed to incorporate auxiliary parameters to aid in reducing uncertainty and improving prediction performance of runway pavement condition data. Using pavement friction data as illustration, the related parameter data included are aircraft landing volume, rainfall, and temperature. This study shows that the proposed ASMI approach provides an analytically meaningful method from pavement engineering point of view that is able to further improve the quality and reliability of imputing missing data for airport pavement management.