Machine Learning-Assisted Probabilistic Creep Life Assessment for High-Temperature Superheater Outlet Header Considering Material Uncertainty

The high-temperature superheater outlet header (Outlet Header) in ultra-supercritical (USC) thermal power plants is subjected to high temperatures and pressures, which increases the risk of creep failure. To assess the structural reliability of the Outlet Header, it is necessary to consider the impact of uncertainty factors. Furthermore, the diverse operating conditions make reliability assessment inconvenient. This study evaluates the creep life reliability of the Outlet Header based on material uncertainty and simplifies the assessment process using machine learning methods. Considering the scatter of creep rupture data, the uncertainty of material constants in the Larson-Miller (LM) model is quantified by randomly sampling a specific number of creep rupture life data. Based on the results of uncertainty quantification and finite element analysis, the distribution of the Outlet Header's creep life is obtained to calculate its reliability under design life. Machine learning is employed to assist in the reliability assessment of creep life under different operating conditions of Outlet Header. The results indicate that Artificial Neural Network (ANN) demonstrates superior performance in this study, and an assessment diagram based on the ANN has been constructed. This approach provides a practical solution for assessing the reliability of high-temperature components in engineering.