Rapid ATR-FTIR-ACE chemometrics for quantifying mineral oil adulteration in synthetic engine lubricants: Comparison...

A rapid, non-destructive chemometric approach using ATR-FTIR spectroscopy coupled with alternating conditional expectation (ACE) regression was developed to quantify mineral oil adulteration in fully synthetic engine lubricants. After spectral preprocessing via Savitzky-Golay filtering and mean-centering, ACE effectively modeled nonlinear spectral-compositional relationships through nonparametric conditional transformations. The method achieved high predictive reliability (R² = 0.983-0.990, Q² = 0.976-0.988, RMSE = 3.325-3.867 (w/w% adulterant)), strong analytical precision (RPD = 6.458-9.256, RER = 25.859-31.211), and a parsimonious model structure (AIC/BIC ≈ 62-75). ACE significantly outperformed PLS-R (Q² = 0.69-0.90) and SVM-R (Q² = 0.64-0.97), demonstrating that ATR-FTIR-ACE provides a robust, interpretable, and computationally efficient framework for high-throughput lubricant authentication and quality control.