Concrete corrosion is a significant threat to the structural integrity of safety-critical sewer systems. Empirical models of corrosion prediction have been developed, but model parameters, such as the pH factor, dissolved sulfide concentration, etc., have high uncertainties, and sewer field monitoring data are limited. This study proposes a Bayesian model-updating method based on observed point-cloud data collected by light detection and ranging inspection. The proposed method introduces a novel likelihood paradigm using the earth mover’s distance to measure the mismatch between observed and predicted data distributions. As the high-dimensional, truncated, and asymmetric characteristics of Bayesian joint posteriors make uncertainty computing difficult, an improved Markov chain Monte Carlo sampler called bounded Hamiltonian Monte Carlo (BHMC) is presented. The effectiveness of the proposed method is demonstrated using a real case. The results show that the established Bayesian framework successfully achieved accurate pipeline reliability estimation and interpreted the critical parameter evolution of concrete corrosion, which contributes to predictive maintenance. Moreover, the superiority of BHMC in computational accuracy and sampling efficiency was validated through comparison with state-of-the-art methods.
Li et al. (Tue,) studied this question.