A comparison of benchtop and micro NIR spectrometers for infant milk formula powder storage time discrimination and particle size prediction using chemometrics and denoising methods

Abstract

This study compared benchtop and micro NIR spectrometers in diffuse reflection mode (1158–2169 nm) to (i) discriminate infant milk formula (IMF) powder samples (n = 500) by storage time (0–12 months) and to (ii) predict IMF powder particle size parameters. Reference particle size parameters were determined using laser-light-scattering. Fast Fourier transform (FFT) and wavelet transform (WT) denoising methods were employed on micro NIR spectra to remove noise and enhance spectral features. Partial least square discriminant analysis (PLS-DA) and partial least squares regression (PLSR) modelling approaches were used to discriminate IMF by storage time and predict powder particle size, respectively. The models developed using the micro NIR denoised spectra had correct identifications (CI p ) of 89–100% for IMF storage time discrimination and RPDPs of 2.29–2.98 for particle size prediction, which were comparable to the model performances achieved using the benchtop NIR spectra (CI p values of 81.3–100% for IMF storage time discrimination and RPDPs of 2.08–3.07 for particle size prediction). This study demonstrated that the potential of micro NIR spectroscopy combined with chemometrics and denoising methods as a suitable process analytical technology (PAT) tool for IMF powder storage time discrimination and particle size prediction along the IMF distribution chain from production to consumer. • Benchtop and micro NIR models can assess IMF powder storage time and particle size. • IMF powder storage time was discriminated using PLS-DA modelling. • IMF powder particle size was predicted using PLSR modelling. • FFT and WT denoising methods are suitable to remove micro NIR spectral noise. • Micro NIR models had overall better performance than benchtop NIR models.