Experimental Study and Prediction of Performance and Emission in an SI Engine Using Alternative Fuel with Artificial Neural Network

Abstract

In this study, the effect of using pure ethanol in different operating conditions of a spark ignition engine was experimentally investigated, and a backpropagation artificial neural network (ANN) model was developed to estimate the engine performance and exhaust emissions. For this purpose, the spark ignition (SI) engine having a compression ratio (CR) of 8.5:1, a single cylinder and air-cooled was used in the engine tests experiments, and the ANN model was created with using the C# programming language. The engine tests were, firstly, conducted for four varied CR, three types of air excess coefficient (AEC) and three different ignition timing (IGT) at 2400 rpm, and the performance and exhaust emission of the engine were recorded. Secondly, the performance and exhaust emission values of SI engine for the same test conditions were estimated with a backpropagation ANN model. The ANN model was trained with the data obtained from the experimental study. The engine torque, brake specific fuel consumption (BSFC), hydrocarbon (HC) emission and carbon dioxide (CO2) emission results obtained from the pure ethanol were compared with those of the gasoline. The comparison was made separately for the same test conditions, and the changes were given as a percentage into the paper. Furthermore, the estimated performance and emission values obtained from the ANN trained with experimental data are compared with verification experiments. According to the results, it can be seen that the percentage change in the difference between the values obtained from the verification experiments, and the estimated results is acceptable. When the results are generally evaluated, it is observed that is improved of the performance and exhaust emissions as using pure ethanol at lowering the IGT and increasing the CR. When all tests are evaluated according to the best results, it is shown that engine torque and BSFC increased respectively about 11% and 51% (20 CA) on average while CO2 and HC emissions reduced about 6% and 49.2% on average. In addition, it is seen that the ANN model can be used to estimate the performance and emissions of an SI engine using alternative fuels.

Prediction of direct carbon emissions of Chinese provinces using artificial neural networks

“…Ye et al (2018) have used neural networks to predict carbon emissions in the construction industry, where the economic development and improvement of standards might significantly impact carbon dioxide emissions in the future [11]. Balki et al (2018) have adopted backpropagation artificial neural networks to develop models that can estimate engine performance and exhaust emissions, finding that carbon emissions of ethanol vehicles are 6% lower than gasoline ones [12]. Liu et al (2017) have combined the chaos theory with the Backpropagation Neural Network (BPNN) to fit and predict carbon emission time series without considering other factors, proposing a method that is easier and more accurate than other prediction methods [13].…”

A Comparative Study of Ethanol and Methanol Addition Effects on Engine Performance, Combustion and Emissions in the SI Engine

“…Balki et al, in their study, have used pure ethanol as fuel, and compared torque, bsfc, HC and CO2 values with that of gasoline. They reported that with using of gasoline as fuel, max torque was 11.1 Nm, min bsfc was 278.8 g/kWh, max HC emission was 196 ppm, max CO2 was 13.7%, on the other hand, with pure ethanol, max torque was 11.41 Nm, min bsfc was 453.7 g/kWh, max HC emission was 115 ppm, max CO2 was 13.2% [19].…”

Prediction of Direct Carbon Emissions of Chinese Provincial Residents under Artificial Neural Networks in Deep Learning Environment

“…Balki et al (2018) used back-propagation artificial neural networks to develop models that can estimate engine performance and exhaust emissions. It was found that the carbon emissions of ethanol vehicles were reduced by 6% compared to gasoline vehicles [12]. Liu et al (2017) used chaos theory combined with BP (Back Propagation) neural network to fit and predict carbon emission time series without considering other factors.…”