Bayesian-Network-Based Evaluation for Corrosion State of Reinforcements Embedded in Concrete by Multiple Electrochemical Indicators

Abstract

The electrochemical indicators including corrosion potential (E_corr), concrete resistivity (ρ), corrosion current density (i_corr), and polarization resistance (R_ρ) are pivotal in the evaluation of the degradation state of reinforcements embedded in concrete. Notwithstanding, extensive investigations traditionally hinge on a singular electrochemical metric for the appraisal of rebar corrosion. The current study transcends this conventional approach by integrating multiple electrochemical detections, significantly improving the accuracy in ascertaining the corrosion status of reinforcing bars within concrete. In this paper, a Bayesian network model is developed, synthesizing results from four electrochemical indictors obtained from published literatures. This model effectively addresses the challenge of integrating unmeasured electrochemical parameters in cases where only a limited set is tested in practical engineering, culminating in a more comprehensive assessment dataset. Further, this study progresses to quantitatively assess the reinforcement corrosion status by devising and fine-tuning an integrated model. The Bayesian network notably excels in extrapolating untested results and accurately determining the thresholds for rebar corrosion status, thus significantly improving the overall assessment capability. The Bayesian network, as employed in this study, computes median E_corr and i_corr values at -282mV and 0.168µA/cm², respectively. These computed values exhibit a deviation within 15% of experimental data, aligning with the uncertainty range stipulated by the ASTM C876-91 standards.