Article: Multi-objective micro-milling parameter optimisation and surface prediction via migration learning Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.52 pp.1 - 19 Abstract: In this study, a multi-objective optimisation framework incorporating migration learning is proposed with the aim of efficiently optimising micro-milling parameters and accurately predicting surface roughness. First, a deep neural network (DNN)-based surface roughness prediction model is constructed as a base model. Subsequently, the pre-trained model is fine-tuned (fine-tuning) using a limited amount of micro-milling experimental data in the target domain to quickly adapt to the target working conditions and significantly improve the prediction accuracy under small samples. On this basis, the migration learning-enhanced prediction model is integrated with a multi-objective optimisation algorithm (e.g., NSGA-II) to construct an optimisation framework. Experimental results show that relying on the millisecond evaluation capability of the migration learning agent model and the improved search strategy of NSGA-II, the Pareto frontier distribution range is widened by 28% and the frontier convergence speed is improved by 42%. Inderscience Publishers - linking academia, business and industry through research

Title: Multi-objective micro-milling parameter optimisation and surface prediction via migration learning

Authors: Pu Zhang

Addresses: School of Advanced Manufacturing Technology, Guangdong Mechanical and Electrical Polytechnic, Guangzhou, 510550, China

Abstract: In this study, a multi-objective optimisation framework incorporating migration learning is proposed with the aim of efficiently optimising micro-milling parameters and accurately predicting surface roughness. First, a deep neural network (DNN)-based surface roughness prediction model is constructed as a base model. Subsequently, the pre-trained model is fine-tuned (fine-tuning) using a limited amount of micro-milling experimental data in the target domain to quickly adapt to the target working conditions and significantly improve the prediction accuracy under small samples. On this basis, the migration learning-enhanced prediction model is integrated with a multi-objective optimisation algorithm (e.g., NSGA-II) to construct an optimisation framework. Experimental results show that relying on the millisecond evaluation capability of the migration learning agent model and the improved search strategy of NSGA-II, the Pareto frontier distribution range is widened by 28% and the frontier convergence speed is improved by 42%.

Keywords: migration learning; micro-milling; multi-objective optimisation; surface roughness prediction.

DOI: 10.1504/IJICT.2025.151171

International Journal of Information and Communication Technology, 2025 Vol.26 No.52, pp.1 - 19

Received: 11 Jul 2025
Accepted: 29 Aug 2025
Published online: 15 Jan 2026 *

Title: Multi-objective micro-milling parameter optimisation and surface prediction via migration learning

Keep up-to-date