Previsão da vazão mássica em uma máquina frigorífica utilizando redes neurais artificiais

• Ano de defesa: 2022
• Autor(a) principal: Vinicius David Fonseca
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA MECÂNICA; Programa de Pós-Graduação em Engenharia Mecanica; UFMG
• Palavras-chave em Português: Previsão da vazão mássica; Máquinas frigoríficas; Redes neurais artificiais; Engenharia mecânica; Redes Neurais (Computação); Máquinas térmicas
• Link de acesso: http://hdl.handle.net/1843/45082

Measuring mass flow rate in refrigeration systems with flow meters can be expensive when taking into account the cost of the equipment itself and the costs related to installation and maintenance. A model based on Artificial Neural Networks (ANNs) can be used to predict the value of the mass flow, at low cost, through easily observed and measured parameters, like temperatures. Additionally, well-known correlations to calculate parameters that directly influence the mass flow rate can be used as input data for the ANN to improve its accuracy. Within this context, the present study aims to develop a Multilayer Perceptrons (MLP) model to predict the mass flow rate of a refrigeration systems. Later, it is presented an alternative mass flow rate meter, using an ANN model programmed in a microcontrolled circuit with only three temperatures as inputs, that was developed and tests using the software Proteus. To develop the ANN model, experimental data were collected in a refrigeration machine in several operating points. Step disturbances were introduced in the mass flow rate to produce transient data. Two different data set were considered in the training process. The first data set contained only steady-state data and in the second data set there were steady-state plus transient data. The mass flow rate estimated through the ANN presented an average error of 0.79 % when considering steady-state and transient data in the training process, and 0.81 % when considering only steadystate data in the training procedure. In both cases, the average error was smaller than the mass flow meter uncertainty.