Quantum Reservoir Computing for Corrosion Prediction in Aerospace: A Hybrid Approach for Enhanced Material Degradation Forecasting

Predicting material degradation is a key challenge across many industries, as environmental factors such as humidity and temperature drive corrosion and related processes. Quantum machine learning offers a promising path forward, but current approaches face well-known challenges including barren plateaus and measurement overheads. To address these issues, we explore quantum reservoir computing (QRC) for time-series prediction. While QRC can model dynamical systems effectively, designing expressive yet shallow circuits remains difficult. Inspired by the classical onion echo state network (ESN) model, which concatenates smaller reservoirs to capture multiple time scales, we introduce the onion QRC—a quantum analog that evolves multiple small reservoirs in parallel to encode both long- and short-term dependencies while maintaining shallow circuit depth. We show that modifying rotation angles or the number of mid-circuit measurements tunes the eigenvalue spectrum, allowing control over temporal memory. The onion QRC outperforms both a single QRC and a classical reservoir in predicting aluminum alloy degradation under varying environmental conditions, with further improvement achieved by combining it with an additional classical reservoir layer.