Computational Neural Networks for Resolving Nonlinear Multicomponent Systems Based on Chemiluminescence Methods

Abstract

This paper proves that computational neural networks are reliable, effective tools for resolving nonlinear multicomponent systems involving synergistic effects by using chemiluminescence-based methods developed by continuous addition of reagent technique. Computational neural networks (CNNs) were implemented using a preprocessing of data by principal component analysis; the principal components to be used as input to the CNN were selected on the basis of a heuristic method. The leave-one-out method was applied on the basis of theoretical considerations in order to reduce sample size with no detriment to the prediction capacity of the network. The proposed approach was used to resolve trimeprazine/methotrimeprazine mixtures with a classical peroxyoxalate chemiluminescent system, such as the reaction between bis(2,4,6-trichlorophenyl)oxalate and hydrogen peroxide. The optimum network design, 9:5s:2l, allowed the resolution of mixtures of the two analytes in concentration ratios from 1:10 to 10:1 with very small (less than 5%) relative errors.

Time-resolved sensor arrays

“…Up to ternary mixtures have been quantified by use of this single-sensor approach. Hervás et al [5] use a peroxyoxalate chemiluminescent system to resolve mixtures of trimeprazine and methotrimeprazine. Several publications have dealt with quantification of analytes in gas mixtures on the basis of single-polymer sensors [6,7,8].…”

“…Some publications use rather arbitrary descriptors of the sensor signals, for example "most positive change in signal", "steepest positive slope of response", "average change in intensity for all time intervals", and many more [12,13], whereas other publications use the actual sensor signal at fixed time points or time intervals. In most of the latter publications sensor responses are recorded using rather coarse time resolution [5,14], which furnishes an easily manageable quantity of information. If several analytes result in similar or very rapid sensor responses, however, the risk of losing important information necessitates use of a fine time grid, which requires more sophisticated data analysis [8,15].…”

“…For the often more demanding quantitative data analysis widespread partial least squares (PLS) has been used [4,16]. Application of neural networks to time-resolved data is often hindered by too high a number of variables for too few calibration samples; if so, data compression by a combination of PCA and BP-ANN can help [5]. Dieterle et al have suggested two concepts, based on growing neural networks and on combination of genetic algorithms with neural networks, for simultaneous calibration and variable selection for non-linear time-resolved data sets with many variables and a limited number of calibration samples [8,15].…”

Use of Pruned Computational Neural Networks for Processing the Response of Oscillating Chemical Reactions with a View to Analyzing Nonlinear Multicomponent Mixtures

“…Could a simpler alternative be used instead? These questions were answered with the aid of response surface [P ]* ) 1.54ĥ 1 + 0.37ĥ 2 -0.78ĥ 3 (16) [GÂ ]* ) -0.69 -1.74ĥ 1 + 1.89ĥ 2 + 1.22ĥ 3 (17)…”

“…Methods for kinetic multicomponent determinations based on various chemometric tools have proliferated in the past few years . Thus, linear and extended Kalman filtering − and, recently, multivariate calibration techniques based on factor analysis and artificial intelligence such as principal component regression (PCR), , partial least squares (PLS), − and CNNs − have found increasing application in this field. These methods require no prior knowledge of the reaction rates for the species in the analytical system; also they avoid or reduce synergistic effects and other unknown sources of nonlinearity.…”