EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Machine Learning Assisted Electrochemical Prediction of Corrosion Resistance in Additively Manufactured Ti6Al4V Biomaterials under Simulated Physiological Conditions

Samuel Anane Akenten, Onuh Matthew Ijiga

International Journal of Scientific Research in Mechanical and Materials Engineering · 22.05.2026 · 10 (3) · 01-33

DOI: 10.32628/ijsrmme261038 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Scientific Research in Mechanical and Materials Engineering, online erschienen am 22.05.2026. Verfasst von Samuel Anane Akenten und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Additive manufacturing of Ti6Al4V biomaterials has transformed the development of patient-specific orthopedic and dental implants due to its geometric flexibility, reduced material waste, and rapid fabrication capability. However, the layer-wise fabrication process introduces microstructural heterogeneity, residual stresses, anisotropic grain orientation, and localized porosity, all of which significantly influence electrochemical corrosion behavior under physiological exposure. Conventional electrochemical corrosion evaluation methods such as potentiodynamic polarization, electrochemical impedance spectroscopy (EIS), and immersion-based degradation testing remain experimentally intensive, time-consuming, and limited in predictive adaptability under dynamic biomedical conditions. This study proposes a novel machine learning-assisted electrochemical prediction framework termed HMECP-Net (Hybrid Multiscale Electrochemical Corrosion Predictor Network) for accurate prediction of corrosion resistance in additively manufactured Ti6Al4V biomaterials under simulated physiological environments. The proposed framework integrates electrochemical descriptors, microstructural imaging features, and additive manufacturing process parameters into a unified predictive architecture. Experimental data are generated using Ti6Al4V specimens fabricated through Selective Laser Melting (SLM) and Electron Beam Melting (EBM), exposed to simulated body fluid (SBF), phosphate-buffered saline (PBS), and Hank’s balanced salt solution at 37°C. Electrochemical features include corrosion potential (Ecorr), corrosion current density (Icorr), charge transfer resistance (Rct), double-layer capacitance (Cdl), impedance magnitude (|Z|), phase angle response, and pitting initiation potential (Epit). Microstructural descriptors extracted through SEM and XCT analysis include pore density, grain boundary irregularity, surface roughness, oxide thickness, and defect morphology metrics. The proposed HMECP-Net combines Convolutional Neural Networks (CNN) for image-derived defect characterization, Bidirectional Long Short-Term Memory (BiLSTM) for electrochemical temporal signal interpretation, and a novel Adaptive Electrochemical Attention Fusion Module (AEAFM) for feature weighting and multimodal integration. Comparative benchmarking is conducted against Random Forest Regression, Support Vector Regression (SVR), XGBoost, standalone CNN, standalone LSTM, and Gradient Boosting frameworks. Results demonstrate that HMECP-Net achieves superior predictive performance with R² = 0.984, RMSE = 0.0068 mA/cm², MAE = 0.0041 mA/cm², and prediction accuracy of 97.8%, outperforming XGBoost (R² = 0.941), SVR (R² = 0.902), Random Forest (R² = 0.918), and conventional LSTM (R² = 0.956). Comparative graphical analysis includes polarization curve prediction accuracy, Nyquist impedance reconstruction, SHAP-based feature importance ranking, model convergence analysis, runtime-performance benchmarking, and cross-environment generalization performance. The findings establish HMECP-Net as a robust intelligent corrosion prediction framework capable of accelerating biomaterial qualification, reducing experimental burden, and enabling predictive implant durability assessment in biomedical engineering applications.

Zitieren

APA

Akenten, S. A., & Ijiga, O. M. (2026). Machine Learning Assisted Electrochemical Prediction of Corrosion Resistance in Additively Manufactured Ti6Al4V Biomaterials under Simulated Physiological Conditions. International Journal of Scientific Research in Mechanical and Materials Engineering, 10(3), 01-33. https://doi.org/10.32628/ijsrmme261038

Vancouver

Akenten SA, Ijiga OM. Machine Learning Assisted Electrochemical Prediction of Corrosion Resistance in Additively Manufactured Ti6Al4V Biomaterials under Simulated Physiological Conditions. International Journal of Scientific Research in Mechanical and Materials Engineering. 2026;10(3):01-33. doi:10.32628/ijsrmme261038

BibTeX
@article{Akenten2026,
  author = {Akenten, Samuel Anane and Ijiga, Onuh Matthew},
  title = {Machine Learning Assisted Electrochemical Prediction of Corrosion Resistance in Additively Manufactured Ti6Al4V Biomaterials under Simulated Physiological Conditions},
  journal = {International Journal of Scientific Research in Mechanical and Materials Engineering},
  year = {2026},
  volume = {10},
  number = {3},
  pages = {01--33},
  doi = {10.32628/ijsrmme261038}
}
Kostenlos starten →