Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
PERFORMANCE OF COMPRESSION IGNITION ENGINES FUELED BY DIESEL-BIODIESEL-ETHANOL BLENDS USING MODELS BASED ON ARTIFICIAL NEURAL NETWORKS
Submission Author:
Florian Alain Yannick Pradelle , RJ
Co-Authors:
Naiara Rinco de Marques e Carmo, Sergio Braga, Florian Alain Yannick Pradelle
Presenter: Florian Alain Yannick Pradelle
doi://10.26678/ABCM.ENCIT2022.CIT22-0220
Abstract
This work aimed to develop different topologies of Artificial Neural Networks (ANNs) to obtain performance parameters of Internal Combustion Engines (ICE) fueled with blends of diesel – biodiesel – ethanol. Eight targets were evaluated: thermal efficiency, brake specific consumption of the blend and of ethanol, exhaust gas temperature, compression and expansion polytropic coefficients, maximum rate of heat released, and maximum pressure. The possible inputs were fuel consumption, ethanol content, lower heating value (LHV), among others. Different settings (type of network, number of neurons in the hidden layer, transfer function, among others) were considered to have ANNs with a high value of coefficient of determination (R²) for training and test datasets, and smaller errors like Mean Absolute Percentage Error (MAPE). Two scenarios were compared: networks with the significant inputs of the correlation matrix and another with less inputs. For the networks with better performances, response surfaces were plotted for qualitatively analysis. It was possible to obtain models with good representation of almost all the mentioned parameters (obtaining R² values in the ranges of 0.606 – 0.995 for training and of 0.685 – 0.992 for test data). In consequence, the models can be used for optimization of the operation of the studied engine with such blends. Only the polytropic coefficients for compression and expansion could not be modelled properly, because the response surfaces did not represent the expected behavior.
Keywords
INTERNAL COMBUSTION ENGINES, machine learning, Diesel-Biodiesel-Ethanol Mixtures, Efficiency, Combustion
