Bayesian inference of hidden corrosion in steel bridge connections: Non-contact and sparse contact approaches

The work is numerical in nature and involves the solution of stochastic inverse problems, in a Bayesian setting, in order to quantify uncertainty in modeling parameter estimation describing hidden corrosion (i.e. corrosion that cannot be directly visualized). A non-contact approach, involving a digital infrared camera, is described. Additionally, the paper also describes a means for employing so-called Gappy Proper Orthogonal Decomposition (Gappy-POD) in permitting a tiny number of contact temperature sensors to be applied for characterization of the same hidden corrosion.