Comparison of LMP and MHS+MG Models for Oxidation-Corrected Creep Rupture Life Prediction of Grade 23 Steel: English

Reliable creep rupture-life prediction is essential for high-temperature components operating under sustained stress. Grade 23 steel (T23) offers improved creep strength over conventional 2.25Cr–1Mo steels, but long-duration tests in air are affected by oxidation-induced section loss. This study compares unconstrained quadratic Larson–Miller Parameter (LMP), monotonic quadratic LMP and Modified Hyperbolic Sine plus Monkman–Grant (MHS+MG) stress-function models using the same oxidation-corrected Grade 23 dataset. The dataset contains 135 observations across 500, 550, 600 and 650 °C. The unconstrained quadratic LMP gives reasonable overall accuracy but develops an internal non-monotonic turning point, most visibly at 650 °C. Constraining the quadratic LMP to remain monotonic removes this behaviour but reduces predictive accuracy. MHS+MG provides the strongest overall performance under both ordinary least-squares and robust residual-weighted calibration. The ordinary least-squares MHS+MG fit gives RMSE = 0.2614 in log₁₀(tᵣ) space, R² = 0.9363 and MAPE = 0.5082, whereas the robust-weighted fit gives RMSE = 0.2719, R² = 0.9311 and MAPE = 0.5303. The superior MHS+MG performance therefore cannot be attributed solely to the robust fitting strategy. Within this oxidation-corrected dataset, MHS+MG provides the best balance of accuracy, stability and physically consistent stress–life trend behaviour.