Data-Driven Characterization and Prediction of Surface Properties in Plasma-Nitrided AISI 8620 Steel

Abstract

This investigation presents a comprehensive evaluation of surface modification achieved through plasma nitriding of AISI 8620 steel, incorporating both conventional experimental characterization and advanced computational modeling approaches. Plasma nitriding treatments were conducted on AISI 8620 specimens with and without prior carburization-quenching across temperature ranges of 450-520 degrees C and treatment durations of 10-20 h under controlled atmospheric conditions. Comprehensive material characterization encompassed scanning electron microscopy, optical metallography, x-ray diffraction analysis, microhardness profiling, and surface topography assessment. Tribological evaluation was performed using ball-on-disk testing methodology to quantify wear behavior and friction characteristics of the AISI 8620 steel samples. Statistical significance of processing variables was evaluated through analysis of variance (ANOVA) techniques, while machine learning algorithms with Leave-One-Out Cross-Validation were employed to assess model reliability and identify critical parameter relationships for the limited dataset size (n = 12). Computational analysis revealed substantial correlation between processing temperature and diffusion zone development (r = 0.926), with Leave-One-Out Cross-Validation demonstrating that only diffusion layer thickness (LOOCV R2 = 0.958) and surface microhardness (LOOCV R2 = 0.488) models are suitable for predictive purposes. Model validation revealed that surface roughness, white layer thickness, and friction coefficient models showed negative LOOCV R2 values, indicating these relationships should be considered exploratory only. Process parameter optimization is constrained to reliable models within the experimental range (450-520 degrees C, 10-20 h).