The aim of this work is (i) to study the ability of near infrared (NIR) spectroscopy to follow in-line the agglomeration process of durum wheat semolina using mechanical mixer according to two different water supply conditions (i.e. high versus low water flow rates) and to four different water addition levels, and (ii) to describe the associated structural changes of agglomerates. The results showed that analyses of spectral variations can be useful to discriminate different agglomeration processes, which are associated to different kinetic changes in terms of agglomerate size distribution and water content per size fraction. Principal component analysis (PCA) has been used to qualitatively describe physical and chemical variations occurring during the wet agglomeration mechanisms and to identify specific kinetics and characteristic times. Whatever the water supply condition, kinetic variations of principal components (PCs) on raw and second derivative NIR spectra are observed. Characteristic times have been identified on PC scores and an attempt to link these times to changes of agglomerates physical and chemical properties have been carried out, mainly based on the analysis of associated loading spectra. (C) 2013 Elsevier Ltd. All rights reserved.