Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
THE USE OF NEURAL NETWORK FOR DETERMINATION OF THE LOAD AND ADJUSTMENT OF THE INTERNAL TEMPERATURE OF A HOUSEHOLD REFRIGERATOR
Submission Author:
Matheus Mugayar Monteiro , MG , Brazil
Co-Authors:
Matheus Mugayar Monteiro, Stênio Barbosa Caldeira, Hudson Guimarães, Henrique José Agrizzi Altoé, Luiza Cordeiro, Antônio Maia
Presenter: Matheus Mugayar Monteiro
doi://10.26678/ABCM.COBEM2021.COB2021-1330
Abstract
The capacity control method most used today in domestic refrigerators is a thermostat that regulates when the compressor will turn on and off, maintaining the internal temperature within the desired values. However, this control method induces significant energy losses due to transients generated during the compressor start and stop. These losses became even more significant when the number of start/stop events increases due a to diverse range of factors. One of them being the adjustment of the internal temperature defined by the user, which has a directed influence on the number of on/off cycles. This parameter is modified many times without a well-defined criterion, increasing, even more, the energetic losses of the refrigerator. This adjustment could be done via an algorithm, considering the product load and the ideal temperature for food perseveration, seeking to minimize the energy consumption, without the risk of harming the user's health. Therefore, this work aims to develop an Artificial Neural Network (ANN) to estimate the product load in the refrigerator and utilize this information to define the adjustment of the internal temperature of operation, contributing to reduce the number of on/off cycles of the equipment. Experimental tests were carried out to characterize the functioning of the refrigerator with different product loads and ambient temperatures. The neural network will use regression techniques and various parameters were autonomously tested using a tuning algorithm, to ensure that the best result can be achieved. As a result, this algorithm was able to estimate the load within the refrigerator with a precision of ±0,88 kg.
Keywords
Domestic refrigerator, Energy consumption, Artificial neural network (ANN), product load
