An Expectation-Maximization Algorithm Embedded Framework for Safety Assessment of Reinforced Concrete Systems
Vincent Z. Wang and Nagaratnam Sivakugan
Appears on pages(s):
appraisal; compressive strength; EM algorithm; failure; missing data; reinforced concrete; statistics
This paper studies the safety assessment of in-service reinforced concrete (RC) structures with incomplete appraisal data. An appraisal data missingness event may occur under a series of circumstances, such as the breakdown of the data acquisition system involved. To deal with the incomplete appraisal data, two straightforward options—that is, pairwise and listwise parameter estimation schemes—are first explored, and the discussion focuses on the positive-semidefiniteness criterion, the additional information loss, and the probability of noninformativeness. An expectation-maximization (EM) algorithm embedded safety assessment framework for in-service RC systems is then formulated as a remedy for the missingness of the appraisal data. Through a designed statistical hypothesis testing scheme, the efficacy and the general applicability of the framework are illustrated using both the numerically simulated appraisal data and the data collected by a rebound hammer during a field investigation. The proposed framework has the potential to be incorporated into the sustainable RC structural design paradigms.