The industrial manufacture of membranes is well established at the present time. More than any other process, the production of ultrafiltration membranes by immersion or evaporation (phase inversion) precipitation techniques is one of the most common. In many cases, the macroscopic properties of the membranes are similar from one membrane to another and it is impossible to distinguish them, while they differ notably from the microscopic point of view. The aim of this work is to develop a method for the classification of polysulfone ultrafiltration membranes prepared either by immersion or evaporation. It also presents a classification of those membranes by thickness. For these purposes, near infrared spectroscopy (NIR) combined with chemometric techniques are attempted here for the first time in the area of rnembrane research. The NIR technique permits fast analytical measurement of membrane samples, together with the possibility of characterization in on-line mode, without destruction or invasion of the samples. This appears to be an excellent routine analysis for purposes of membrane classification. The membranes were prepared in our Universitat Autonoma de Barcelona (UAB) laboratory and, after obtaining the NIR spectra, principal component analysis (PCA) was used to describe the system. The second stage involved the application of a pattern recognition method: supervised independent modeling of class analogy (SIMCA) in order to classify unknown samples. Finally, the ultrafiltration membranes were classified in terms of the membrane preparation technique (immersion or evaporation). In addition, membranes prepared by immersion were classified by thickness. (C) 2007 Elsevier B.V. All rights reserved.