Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
MASS FLOW PREDICTION IN A REFRIGERATION MACHINE USING ARTIFICIAL NEURAL NETWORKS
Submission Author:
Vinícius Fonseca , MG
Co-Authors:
Vinícius Fonseca, Antônio Maia
Presenter: Vinícius Fonseca
doi://10.26678/ABCM.COBEM2021.COB2021-0763
Abstract
In the works related to mathematical modeling of refrigeration systems, the correct estimation of mass flow rate is essential. This parameter can also be used to monitor the system and evaluate aspects related to performance and system maintenance. Measuring with flow meters can be expensive taking into account the cost of the equipment itself and the costs related to installation and maintenance. Alternatively, mathematical correlations can be used to estimate the mass flow rate. Considering this last approach, 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 pressures and 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. The test bench consists of a refrigeration machine, operating with R134a as the primary fluid and pure water as the secondary fluid in the evaporator and condenser. Experimental data was collected for several different permanent and transient regimes. Step disturbances were introduced in the mass flow rate to produce data during the transient response. Two training cases were considered, with only the steady-state data and with both, steady-state and transient response The mass flow rate estimated had maximum error of 4.71 % with all data in the training and 3.81 % with only the steady-state data in the training.
Keywords
Mass flow prediction, Refrigeration machines, Artificial neural networks
