Corrosion-induced cracks represent a distinctive measure of damage in existing reinforced concrete (RC) structures. In this paper, a Bayesian updating framework is established to inform corrosion exposure scenarios based on the distribution of crack widths in RC beams. The diffusion process of chlorides in concrete members is simulated based on tri-dimensional probabilistic cellular automata algorithm. The formation of concrete damage due to the expansion of steel corrosion products is modelled by means of a finite element formulation of nonlinear structural analysis under uncertainties. The Bayesian model with structural reliability methods is adopted to updated the corrosion exposure scenario based on the distribution and characteristics of concrete crack widths though an inverse problem. The proposed framework is validated based on the results of experimental laboratory tests on corroded RC beams available in literature.