Machine learning based optimization of titanium electropolishing using artificial neural networks and Taguchi design in eco-friendly electrolytes

• Published in: Scientific reports Vol. 15; no. 1; pp. 28561 - 18
• Main Authors: Hwang, Hyun-Kyu, Kim, Seong-Jong
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 05.08.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 639/166; 639/638/161; Artificial neural network; Corrosion resistance; Deep learning; Design of experiments; Distilled water; Electrolytes; Electropolishing; Epoxy resins; Ethylene glycol; Humanities and Social Sciences; Medical equipment; multidisciplinary; Neural networks; Optimization; Roughness; Science; Science (multidisciplinary); Sustainable production; Taguchi robust design; Titanium; Voltage; Titanium; Artificial neural network; Taguchi robust design; Electropolishing; Roughness
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-09416-x

This research proposes an integrated optimization framework combining artificial neural networks (ANN) and Taguchi robust design for an eco-friendly deep eutectic solvent-based electrolyte. Five key process parameters—applied voltage, processing time, temperature, electrolyte composition (choline chloride to ethylene glycol ratio) and distilled water concentration—were systematically analyzed for their effects on the surface roughness of electropolished titanium. To address data scarcity and improve the generalization performance of the model, Gaussian noise with a mean of 0 and a standard deviation of 0.05 was applied to the input variables to augment the dataset. The multilayer perceptron-based ANN model effectively learned the nonlinear interactions between process parameters and achieved a high predictive accuracy with a coefficient of determination (R 2 ) of 0.981. The optimal process conditions derived from the ANN model were 20 V applied voltage, 21 min of processing time, a 1:4 electrolyte ratio, 0% distilled water and 42℃, under which a minimum surface roughness of 4.162 nm was experimentally confirmed. This investigation provides a practical pathway for optimizing processes in environmentally sustainable manufacturing by quantitatively analyzing the nonlinear complexity of electrochemical surface treatments.