Rate constants and absorption coefficients from experimental data: An inversion procedure based on recursive neural networks

A general procedure to solve chemical kinetics inverse problems based on recurrent neural networks is discussed in this work. As a first application, rate constants are calculated from the product concentration for the hydrolysis mechanism of 2,7-dicyanonaphthalene molecules. In a second analysis, rate constants and absorption coefficients are obtained from ultraviolet absorbencies data. The present method efficiency is compared with the Simplex and Levenberg–Marquardt methods, commonly used in nonlinear regression techniques. The approach is simple, numerically stable and robust with respect to errors in the initial conditions or experimental data.