A numerical study with surrogate-assisted multi-objective optimization of cross-flow liquid cooling for cylindrical Li-ion cell array

• A CFD-ML framework has been developed for a cross-flow liquid-cooled 20 × 18650 BTMS. • Transient conjugate simulations demonstrate safety thresholds (T max < 308 K, ΔT < 5 K). • Log-scaled second-order surrogate appears to predict ΔT and ΔP well with high accuracy. • Selection of Pareto knee-point yields compact design while maintaining performance. • The selected optimized design reduced temperature non-uniformity by 69.7% relative to the no-cooling baseline.