ADAPTIVE RESONANCE THEORY-BASED NEURAL-NETWORK FOR SUPERVISED CHEMICAL-PATTERN RECOGNITION (FUZZYARTMAP) .2. CLASSIFICATION OF POSTCONSUMERPLASTICS BY REMOTE NIR SPECTROSCOPY USING AN INGAAS DIODE-ARRAY

The supervised working FuzzyARTMAP pattern recognition algorithm has b een applied to automated identification of post-consumer plastics by n ear-infrared spectroscopy (NIRS). Experimentally, a remote operating p arallel multisensor device, based on a rapid InGaAs diode array detect or combined with new collimation optics, has been used. The laboratory setup allows on-line identification of more than 100 spectra per seco nd. Internal parameter settings of FuzzyARTMAP were varied to explore their influence on the classifier's behavior. Discrimination results o btained were better than those from an optimized multilayer feedforwar d backpropagation artificial neural network (MLF-BP) and significantly better than those provided by the partial least squares method (PLS2) . Additional advantages of FuzzyARTMAP compared to these two classifie rs are a significantly higher speed of calibration, the chemical inter pretability of network weight coefficients and a built-in detector aga inst extrapolations.