Desarrollo de un sistema de control inteligente basado en redes neuronales para el proceso de pasteurización de leche en la Planta de Lácteos ESPOCH.

Abstract

An intelligent control system based on neural networks for the pasteurization process of milk in the Dairy Plant ESPOCH was developed which allows to obtain pasteurized milk, and produce 3 types of cheese; fresh, ripe or Andean and mozzarella or acid. A mechanical-structural design of the pasteurizer was developed, the internal circuitry, implementation of intelligent control system, control panel and arrangement of the elements used, the application allowed to verify the operation of the system. The neurocontroller system as well as the Human Machine Interface (HMI) was made in the engineering program Matlab 2014 with the aim of automating and optimizing the pasteurization process slow and open method (VAT), it had been doing manually without proper safety standards both for operators and for the raw material, two artificial neural nets (ANN) one for heating and one for cooling the milk to avoid incongruity of data and allow the system to converge 100%. Uptake of temperatures, the activation of different actuators used, is performed using the serial communication between the Arduino Mega 2560 and PC card containing the 2014 Matlab program. By analyzing data and functional tests it was determined that the intelligent control system based on neural networks type feedforward backpropagation whit Levenberg Marquardt and function regularization (MSE) can be used in processes of industrial type where uses large amounts of data. The use of the control system based on ANN has reduced operator handling up to 80% within the system. It is recommended to follow the sequence of steps set in the HMI so that no failures in income, training and simulation system of ANN.