Simulación y validación del proceso de esterificación no catalítica de ácidos grasos libres que componen el aceite karanja

Abstract

The objective of this work is to simulate and validate a non-catalytic esterification process of Free Fatty Acids (FFA), components of karanja oil (Pongamia pinnata), in order to process the data that will feed an Artificial Neural Network (ANN) which will estimate the mole fractions of the most relevant compounds in the system. The ANN was designed in the MATLAB program, based on the sample of one hundred data generated by a previously validated simulation and subjected to a sensitivity analysis in the DWSIM Software, which revealed the independent and dependent variables of the proposed sequence. The ANN inputs were: mole fraction of water from Feeding stream 1, the mole fraction of oleic acid from the Oil stream, conversion percentage of chemical reaction, and pressure drop in the reactor; the following variables correspond to the outputs: the mole fraction of methyl oleate, the molar fraction of methanol, the mole fraction of triolein and the mole fraction of trilinolein of the liquid stream; the mole fraction of methanol and the mole fraction of water from the gas stream. By using the Bayesian regularization algorithm, together with thirty hidden neurons, it was possible to visualize a mean square error (MSE) and a total regression coefficient (R) of 0.00000411 and 0.99, respectively. A statistical analysis confirmed a 95% of reliability, the adequate predictive capacity of the network. From the simulation, 1.21 kmol/h of methyl oleate was produced and from the residual methanol it was possible to collect 99.75% in the gaseous state. It is recommended to extend ANN learning, inspecting new input and output conditions to create a powerful and complete tool that supports the development of industrial processes that implement oils of residual origin.