Comprehensive framework for the hyper-robust Bayesian calibration of a nondestructive testing instrument

摘要

Reinforced concrete structures degrade due to rebar corrosion, leading to the loss of steel rebar volume. Nondestructive techniques (NDT) are employed for knowing the current state of a corroding structure. However, there are uncertainties in the reported values because of measurement error and noise. In this paper, probabilistic measurement error models (MEM) - relating the "true" value to the NDT reported value - are used, and instrument calibration is proposed to quantify the systematic error and random error. Additionally, there is uncertainty in choosing an appropriate MEM from an ensemble of possible models. This entire process is proposed here within a Bayesian framework of uncertainty quantification. A hyper-robust calibration approach is adopted to avoid the calibrated measurement's sensitivity to the uncertainties in MEM parameters and in the model selection, such that the actual estimate of the corrosion rate is quantified in the form of a probability density function. The proposed procedure is demonstrated for a linear polarisation resistance based instrument. The proposed approach is found to be suitable for the study case and also general enough to be applied to other NDT instruments.