An artificial immune system for the identification of polymer materials based on near infrared (NIR) spectra (in English)
Polimery 2014, No 7-8, 575
This work presents an immune algorithm elaborated for the identification of polymers based on their NIR spectra. It uses the mechanisms and rules typical of natural immune systems. The identification of a polymer consists of a comparison of its NIR spectrum with reference spectra contained in a database. The algorithm acts in two stages. In the first stage, it compares the positions of the main absorption bands in the investigated spectrum with those of spectra from the database. Based on this comparison, the most similar reference spectra are selected. In the second stage, the shape of the numerical derivative of the investigated spectrum is compared with the shapes of the derivatives of the reference spectra selected in the first stage. Our investigations have shown that the algorithm is very effective and reliable. The algorithm can be used both for the identification of polymers in large databases and for the protection of such databases from an uncontrolled expansion.