Publication: Artificial intelligence-based optimization of pack carburizing parameters for low carbon steel with different wood charcoals using AdaBoost-GRNN
0
0
Issued Date
2025-01-01
Resource Type
ISSN
02683768
eISSN
14333015
Scopus ID
2-s2.0-105016802170
Journal Title
International Journal of Advanced Manufacturing Technology
Rights Holder(s)
SCOPUS
Bibliographic Citation
International Journal of Advanced Manufacturing Technology (2025)
Suggested Citation
Wiangkham A., Ariyarit A., Aengchuan P., Homjabok W., Wattikornsirikul N., Koptonglang A., Peanjunat C., Duekunthod R., Teegow R., Thammachot N. Artificial intelligence-based optimization of pack carburizing parameters for low carbon steel with different wood charcoals using AdaBoost-GRNN. International Journal of Advanced Manufacturing Technology (2025). doi:10.1007/s00170-025-16508-5 Retrieved from: https://hdl.handle.net/20.500.14740/50555
Corresponding Author(s)
Other Contributor(s)
Abstract
Carbon steel is a widely used alloy, with mechanical properties heavily influenced by its carbon content and heat treatment methods. Among these, Surface hardening techniques like pack carburizing are prominent for enhancing wear resistance while Maintaining a tough core. This process involves diffusing carbon from solid carbonaceous Materials, typically charcoal, into low-carbon steel. This study explores the optimization of the pack carburizing process for AISI 1015 low-carbon steel using sustainable materials, namely various wood-based charcoals (eucalyptus, bamboo, tamarind, cassava rhizome, and coconut shell) and limestone as an energizer. The investigation focused on the effects of charcoal type, compound form (powder or granular), and experimentally investigated carburizing time to assess their effects on the surface carbon content and hardness. As results demonstrate, eucalyptus charcoal, particularly in powdered form, showed superior performance due to its high carbon content and effective diffusion characteristics. Besides experimental investigation, an Artificial Intelligence-based optimization procedure was employed. A multi-objective optimization framework was developed using the Non-dominated Sorting Genetic Algorithm II (NSGA-II), with the objective function modeled via an AdaBoost ensemble, combined with a generalized regression neural network (GRNN), well-suited for small data sets. The model demonstrated high predictive accuracy (average R<sup>2</sup> > 0.93 while MAPE < 6%). SHAP analysis revealed charcoal type and carburizing time as dominant factors, excluding the factor of Surface depth that normally affects the results. Optimal parameters achieved high carbon enrichment and hardness at a depth of 0.2 mm, carburized within 3.80–4.11 h through eucalyptus charcoal powder compound. This integrated approach supports the development of cost-effective, eco-friendly surface hardening processes with enhanced material performance.