AI model identifies electrolyte conditions to reduce aluminium corrosion in batteries

Researchers at the University of Waterloo have developed an artificial intelligence-based framework to analyse aluminium corrosion and identify electrolyte conditions that could improve the performance of aluminium-air batteries.

Aluminium-air batteries have attracted interest as a potential energy source for electric vehicles because aluminium is abundant, recyclable and relatively inexpensive. However, corrosion of the aluminium anode in alkaline electrolytes remains a major challenge, as it wastes active material, generates hydrogen gas and affects battery reliability.

Published in Frontiers in Batteries and Electrochemistry, the study examined how a commercial aluminium-magnesium-silicon (Al-Mg-Si) alloy corrodes in sodium hydroxide (NaOH) and potassium hydroxide (KOH) solutions at concentrations of 1 to 10 molar.

Join the fourth edition of webinar "Hedging for recyclers - Become an expert in 6 hours" by Jorge Dyszel.

The researchers used electrochemical tests, X-ray diffraction and an artificial neural network to study the alloy’s corrosion behaviour. They also used two genetic algorithm-based optimisation methods to identify electrolyte conditions that could reduce corrosion.

The experiments showed that corrosion increased with hydroxide concentration at low-to-moderate alkalinity. At higher concentrations, factors such as electrolyte viscosity, the accumulation of dissolved aluminium compounds, and hydrogen bubbles covering parts of the metal surface limited the reaction.

The study also found that the two electrolytes affected the aluminium alloy differently. At the same concentration, KOH showed higher corrosion current densities and more negative corrosion potentials than NaOH, indicating higher corrosion under the test conditions. The researchers linked this difference to how potassium and sodium ions interact with the aluminium surface.

Explore a wide range of aluminium scrap and recycled aluminium products from global suppliers on AL Biz.

The impedance spectroscopy supported these findings, showing lower charge-transfer resistance in KOH than in NaOH at equivalent concentrations. X-ray diffraction detected no crystalline aluminium hydroxide, oxyhydroxide or aluminate corrosion products, suggesting that the corrosion products were mainly amorphous, nanocrystalline or highly hydrated.

AI-based optimisation identifies an operating condition

The team trained an artificial neural network using the experimental data and evaluated 351 model configurations. The best-performing model achieved coefficients of determination (R²) above 0.999 for both corrosion potential and corrosion current density.

The researchers then applied two genetic algorithm approaches to identify an optimal balance between the two corrosion indicators. Both methods identified a KOH concentration of approximately 1.35 molar as the optimal condition for balancing the two corrosion indicators within the model. 

Showcase your brand, aluminium recycling initiatives and sustainability vision in our upcoming magazine: Sustainability & Recycling: Aluminium’s Commitment 2026.

The researchers think the approach could make corrosion testing more efficient by combining laboratory experiments, neural networks and optimisation methods. The framework could also support the development of digital-twin platforms to design and improve next-generation aluminium-based energy systems while reducing corrosion losses. As demand for alternative battery technologies grows, this approach could help researchers explore the potential of aluminium-based energy storage.