Rare-earth ion concentration diagnostic method based on adaptive peak-valley differential UV-Vis absorption...

The global energy transition has driven rare-earth demand beyond supply, creating an urgent need for intelligent separation processes, for which highly selective online monitoring is a key enabler of fine process control and cost-efficient operation. Ultraviolet-visible (UV-Vis) spectroscopy offers non-invasive and rapid response capabilities; however, in solvent-extraction engineering its quantitative accuracy is severely compromised by scattering-induced turbidity as well as baseline and wavelength drift. To address these challenges, this study proposes an ion-concentration diagnostic method that integrates adaptive peak-valley differencing with robust regression: dynamic local differencing is employed to suppress broadband background components, and hierarchical, element-specific modeling is implemented based on spectral characteristics (second-order correction or ridge regression). Optimization results show that, compared with the conventional single-wavelength approach, the proposed strategy substantially mitigates background interference, improving the coefficient of determination (R²) for the weak-signal element Tb from 0.98299 to 0.99971. Incorporating ridge regression further suppresses noise-induced overfitting, raising the R² values of Dy, Pr, and Tb to above 0.999, reducing the limits of detection (LOD) to 0.01275, 0.00200, and 0.3773 g L^-1, and lowering the coefficients of variation (CV) in low-concentration repeat measurements to 2.9%, 1.1%, and 9.98%, respectively. During a four-week, cross-period monitoring test, the spectral feature correlation coefficient consistently remained at an exceptionally high level (r > 0.9999). Overall, the proposed diagnostic method demonstrates strong robustness for micro-level detection and long-term operation, providing a reference technical method for online concentration diagnosis and performance optimization in rare-earth solvent-extraction processes.