Resumen

SANCHEZ ESCALONA, Andres Adrian; GONGORA LEYVA, Ever  y  CAMARAZA MEDINA, Yanan. Neural network with dynamic training for simulation of a heat exchangers system. Energética [online]. 2020, vol.41, n.1  Epub 01-Ene-2020. ISSN 1815-5901.

This research proposed an artificial neural network model with dynamic training in order to predict both fluids outlet temperatures on a monoethanolamine heat exchangers system, doing the training, validation and testing with 31 680 data points gathered through the passive experimentation method. The 4-3-2 multilayer perceptron achieved correlations above 98.76 %, and it was corroborated that dynamic training strategy provided more accurate results than single training. Mean absolute errors of 0.419 and 0.372 K were obtained when applying the first approach (for rich amine and lean amine outlet temperatures as output variables, respectively), as compared to 2.214 and 1.181 K when implementing the second one. Since calculated deviations have no significant implication on the technological process under analysis, proposed model is considered appropriate for simulation of the studied heat exchangers performance.

Palabras clave : heat exchange; monoethanolamine; artificial neural network; modeling; training.

        · resumen en Español     · texto en Español     · Español ( pdf )