In this research, degradation of the metal corrosion for a combined-cycle power plant boiler is investigated. By application of the developed methodology, the corrosion lifetime is estimated for the boiler tubes. At first, with special focus on the corrosion failures, the important failure modes and mechanisms are evaluated for the metallic boiler tubes via FMEA method. It was found that pitting corrosion is one of the most common modes of corrosion. So far, most studies have estimated the lifetime of pitting corrosion using the deterministic data, in which the results are valid only for certain limited conditions. Also, majority of the previous works on the corrosion were conducted experimentally and lesser efforts have been made in the field of the corrosion modeling. The proposed models are based on deterministic approach, however corrosion has a stochastic nature, and it is affected by many stochastic factors. In order to improve deficiencies of available deterministic models, in the present paper, stochastic and uncertainty methods are employed to study the corrosion. The temporal process of metal degradation is analyzed in different conditions through the stochastic approach. A proper degradation model is selected for providing the best estimation of pitting corrosion life. Uncertainty intervals/distributions are determined for some of the model parameters and the parameter intervals are specified. The deterministic model is converted to a probabilistic model by taking to account the variability of the uncertainty input parameters. The model is simulated using Monte Carlo method via simple sampling from the uncertainty distribution of the sources. In order to have better estimation of the parameters and also considering the experimental data in the modeling, the result of the life estimation is updated by the Bayesian method. Finally for the element that is subjected to the pitting corrosion degradation, the distribution of the life estimation is obtained. The uncertainty associated with the result is estimated as the probability of occurrence of each event. Modelling results shows that pitting corrosion has stochastic behavior and the lognormal distribution is the most appropriate option for the pitting corrosion modeling. In order to validate the results, the estimations were compared with the power plant field failure data.

This content is only available via PDF.
You do not currently have access to this content.