Prediction of Iron Wear Metal Concentration in Used Engine Oils from FT-IR Spectra Using Partial Least Squares Regression

Wear metal monitoring is an important component of lubricant condition monitoring but commonly relies on elemental techniques such as inductively coupled plasma optical emission spectroscopy (ICP-OES), which require dedicated laboratory infrastructure and sample preparation. This study evaluates whether Fourier-transform infrared (FT-IR) spectra of used engine oils can be combined with partial least squares (PLS) regression to provide a rapid screening estimate of iron (Fe) concentration. Used petrol and diesel engine oil samples were analyzed by FT-IR spectroscopy and ICP-OES. PLS models were developed using processed FT-IR spectra as predictor variables and ICP-OES-derived Fe concentrations as response variables. For petrol used oil samples, the optimized model employing 18 latent variables achieved a root mean squared error of 5.02 ppm and a coefficient of determination of 0.97 between measured and predicted Fe concentrations. Model loadings indicated contributions from spectral features associated with soot, oxidation, nitration, antioxidant (AO) depletion, and zinc dialkyldithiophosphate depletion. Combining petrol and diesel samples in a single model reduced predictive performance and increased uncertainty, indicating that their differing degradation pathways cannot be adequately represented by one common latent variable model. The approach does not directly measure Fe and is not intended to replace elemental analysis. Instead, it provides a rapid, low-cost screening tool for identifying samples with potentially elevated wear metal concentrations and prioritizing them for confirmatory analysis.