Book Description
Bayesian models are developed to calibrate the accuracies of high-resolution in-line inspection (ILI) tools for sizing metal-loss corrosion defects and to characterize the growth of individual defects on energy pipelines. Moreover, a methodology is proposed to evaluate the time-dependent system reliability of a segment of a pressurized pipeline containing multiple active corrosion defects. The calibration of ILI tools is carried out by comparing the field-measured depths and ILI-reported depths for a set of static defects. The measurement error associated with the field-measuring tool is found to be negligibly small; therefore, the field-measured depth is assumed to equal the actual depth of the defect. The depth of a corrosion defect reported by an ILI tool is assumed to be a linear function of the corresponding field-measured depth subjected to a random scattering error. The probabilistic characteristics of the intercept and slope in the linear function, i.e. the constant and non-constant biases of the measurement error, as well as the standard deviation of the random scattering error are then quantified using the Bayesian methodology. The proposed methodology is able to calibrate the accuracies of multiple ILI tools simultaneously and quantify the potential correlations between the random scattering errors associated with different ILI tools. The corrosion growth model is developed in a hierarchical Bayesian framework. The depth of the corrosion defects is assumed to be a power-law function of time characterized by two power-law coefficients and the corrosion initiation time, and the probabilistic characteristics of the parameters involved in the growth model are evaluated using Markov Chain Monte Carlo (MCMC) simulation technique based on ILI data iv collected at different times for a given pipeline. The model accounts for the constant and non-constant biases and random scattering errors of the ILI data, as well as the potential correlation between the random scattering errors associated with different ILI tools. The model is validated by comparing the predicted depths with the field-measured depths of two sets of external corrosion defects identified on two in-service natural gas pipelines. A simulation-based methodology is proposed to evaluate the time-dependent system reliability of a segment of a pressurized pipeline containing multiple active metal-loss corrosion defects. The methodology considers three distinctive failure modes, namely small leak, large leak and rupture, and incorporates the hierarchical Bayesian power-law growth model for the depth of individual corrosion defect. Both the conventional Monte Carlo simulation and MCMC simulation techniques are employed in the methodology to evaluate the failure probability. The methodology is illustrated using a joint of an underground natural gas pipeline that is currently in service.