Abstract: To improve the applicability and prediction accuracy of pipeline corrosion degradation analysis model and strengthen pipeline safety management, a general Bayesian-based IG process degradation analysis method considering random effect was proposed. According to the random effect information of system degradation and the method of directly processing online prediction dataset, a simple IG model and three IG models with random effects were established by Bayesian method. Based on posterior samples generated by relevant model parameters, the probabilities obtained by Bayesian goodness test were 0.9813 for simple IG process model, 1.00 for Random Drift(RD)-IG model, 0.925 for Random Volatility(RV)-IG model, and 0.9947 for Random Drift-Volatility(RDV)-IG model. The results show that the RD-IG model fits well with data, that the Bayesian analysis method is flexible and has better effect on degradation prediction in the real-time monitoring scenario, and that the degradation analysis considering random effect can improve the accuracy of prediction.

Key words: corroded pipeline, inverse Gaussian (IG), random effect, Bayesian method, degradation model, Markov chain Monte Carlo(MCMC)