Efficiently designing lightweight alloys with combined high corrosion resistance and mechanical properties remains an enduring topic in materials engineering. Due to the inadequate accuracy of conventional stress-strain machine learning (ML) models caused by corrosion factors, a novel reinforcement self-learning ML algorithm combined with calculated features (accuracy R² >0.92) is developed. Based on the ML models, calculated work functions and mechanical moduli, a Computation Designed Corrosion-Resistant Al alloy is fabricated and verified. The performance (elongation reaches ∼30 %) is attributed to the H trapping Al-Sc-Cu phases (-1.44 eV H⁻¹) and Cu-modified η/η' precipitates inside the grain boundaries (GBs).

This paper develops a novel approach for online Lithium-ion (Li-ion) battery model identification and state of charge (SOC) estimation. To account for the SOC-dependent battery dynamics and the static nonlinearity between the open-circuit voltage (OCV) and SOC, we formulate a grey box nonlinear state-space model, in which elements depend on SOC in a polynomial way. For model identification, we propose an online concurrent state and parameter estimation by alternating the recursive least squares algorithm and particle filter; the SOC is computed via Coulomb counting during the modeling. The identified grey box model is then applied for SOC estimation using the particle filter. Simulation with real-world battery measurements demonstrates the effectiveness of the model structure and the estimation approach, which is reflected in accurate terminal voltage estimation and nonlinear OCV-SOC relation, and superior performance regarding SOC estimation compared to state-of-the-art approaches.