Comparison of an adaptive resonance theory based neural network (ART-2a) against other classifiers for rapid sorting of post consumer plastics by remote near-infrared spectroscopic sensing using an InGaAs diode array

Abstract

An Adaptive Resonance Theory Based Artificial Neural Network (ART-2a) has been compared with Multilayer Feedforward Backpropagation of Error Neural Networks (MLF-BP) and with the SIMCA classifier. All three classifiers were applied to achieve rapid sorting of post-consumer plastics by remote near-infrared (NIR) spectroscopy. A new semiconductor diode array detector based on InGaAs technology has been experimentally tested for measuring the NIR spectra. It has been found by a cross validation scheme that MLF-BP networks show a slightly better discrimination power than ART-2a networks. Both types of artificial neural networks perform significantly better than the SIMCA method. A median sorting purity of better than 98% can be guaranteed for non-black plastics. More than 75 samples per second can be identified by the combination InGaAs diode array/neural network. However, MLF-BP neural networks can definitely not extrapolate. Uninterpretable predictions were observed in case of test samples that truly belong to a particular class but that are located outside the subspace defined by training set.

NIR - Remote sensing and artificial neural networks for rapid identification of post consumer plastics

“…However, the separation of all 5 clusters is more complex and cannot be shown in a single 3d-plot. It has been shown that the complete separation between all 5 classes can be achieved by neural network classifiers [9].…”

“…In order to chose the optimal detector material, a detailed preliminary study covering the complete NIR spectral region from 1000 to 2500 nm was carried out [9]. In particular the separation power of different wavelengths was investigated.…”

“…This integration time permits to record 158 spectra per second. Taking also into account the computation time of a neural network based decision algorithm as developed by Wienke et al [9], 75 complete identification cycles can be performed per second. With this algorithm, a reliable distinction between the 5 major types of plastic contained in household garbage (PE, PET, PP, PS and PVC) was shown to be possible.…”

“…Further improvement of the identification system and it's application to real world garbage like entire bottles or plastic bodies with dirt or labels as well as it's adaptation to a moving belt are in progress. A detailed investigation of different pattern recognition techniques has been carried out [9] and will be further extended.…”

ART 2-A for Optimal Test Series Design in QSAR

“…16,[18][19][20][21][22][35][36][37][38][39] 3. EXPERIMENTAL SECTION ART 2-A was applied on a data matrix of 103 aliphatic substituents ( Table 1) described by the hydrophobic constant for aliphatic substituents Fr, H-bonding acceptor (HBA) and donor (HBD) abilities, the molar refractivity (MR), and the inductive parameter F, respectively.…”