Raman spectroscopy in combination with chemometrics was explored as a rapid, non-targeted screening method for the detection of milk powder (MP) adulteration using melamine as an example contaminant. Raman spectroscopy and an unsupervised pattern-recognition method, principal component analysis (PCA), allowed for the differentiation of authentic MPs from adulterated ones at concentrations > 1.0% for dry-blended (DB) samples and > 0.30% for wet-blended (WB) ones. Soft independent modelling of class analogy (SIMCA), a supervised pattern-recognition method, was also used to classify test samples as adulterated or authentic. Combined statistics at a 97% confidence level from the SIMCA models correctly classified adulteration of MP with melamine at concentrations ≥ 0.5% for DB samples and ≥ 0.30% for WB ones, while no false-positives from authentic MPs were found when the spectra in the 600-700 cm-1 range were pre-processed using standard normal variate (SNV) followed by a gap-segment derivatisation. The combined technique of Raman spectroscopy and chemometrics proved to be a useful tool for the rapid and cost-efficient non-targeted detection of adulteration in MP at per cent spiking levels.
Keywords: PCA; Raman spectroscopy; Skim milk powder (SMP); classification; economic adulteration; non-fat dry milk powder (NFDM); non-targeted detection; soft independent modelling of class analogy (SIMCA).