Programme

P.7 -- Posters, VCTP-50

Date: Tuesday, 5 August 2025

Time: 08:30 - 10:00

Tracking Diamond Tool Wear in Iron Machining with Machine Learning Simulation Study

Nguyen Trinh Bao Anh (1), Enriquez John Isaac Guinto (1), Halim Harry Handoko (1), Ogiwara Hiroyuki (2), Michiuchi Masato (2), Oguchi Tamio (3), Morikawa Yoshitada (1)

(1) Graduate School of Engineering, The University of Osaka (2) Sumitomo Electric Industries, Ltd (3) Center for Spintronics Research Network, The University of Osaka

Diamond tools are widely used in ultra-precision machining due to their exceptional hardness and ability to produce high-quality surface finishes. However, when machining ferrous metals like iron, diamond suffers rapid wear, mainly due to complex thermo-chemical reactions at the Fe–C interface. To explore these atomic-scale wear mechanisms, we employed machine learning molecular dynamics (ML-MD) simulations. Our interatomic potential was trained on a dataset of around 6,000 structural configurations and achieved high accuracy, with root-mean-square errors of ~130 meV/angstrom for forces and ~4 meV/atom for energies. The model accurately reproduces structural and energetic properties of pure bcc iron, carbon, and Fe–C systems. We conducted large-scale cutting simulations involving over 8,000 atoms with 1 fs timesteps over nanosecond durations. Results show that using the diamond surface as the clearance face leads to significantly higher wear than when used as the rake face. Lower cutting temperatures were found to reduce wear rates. Among the diamond surfaces tested, diamond(111) showed the highest wear resistance. Despite initially predicting wear rates 2–3 orders of magnitude above experimental values, the results are expected to converge with longer simulation times. This study demonstrates the value of ML-MD in revealing wear mechanisms and guiding tool design in ferrous machining.

Presenter: Nguyen Trinh Bao Anh