Developing a model for prediction of the combustion performance and emissions of a turboprop engine using the long short-term memory method. (15th October 2021)

• Record Type: Journal Article
• Authors: Kayaalp, Kiyas; Metlek, Sedat; Ekici, Selcuk; Şöhret, Yasin
• Is Part Of: Fuel. Volume 302(2021)
• Publication Date: 2021-10-15
• Subjects: Aircraft engine -- Artificial intelligence -- Combustion efficiency -- Exhaust gas emissions -- Long short-term memory; Fuel -- Periodicals; Coal -- Periodicals; Coal; Fuel; Periodicals; 662.6
• DOI: 10.1016/j.fuel.2021.121202
• Languages: English
• ISSNs: 0016-2361

Abstract

Highlights: Determination of the exhaust emission index and combustion efficiency of the single shaft T56-A-15 engine. Estimating the exhaust emission index values and engine combustion efficiency with optimum hyper-parameters. Estimation of CO, CO2, UHC and NO2 exhaust emission parameters with LSTM, which is one of the current machine learning methods. The effect of fuel flow, engine shaft speed and air–fuel ratios on the estimation of exhaust emission indices. Abstract: In this study, the exhaust emissions index and combustion efficiency of the single shaft T56-A-15 engine are modeled using the Long-Short Term Memory (LSTM) method, one of the recent artificial intelligence algorithms. For this purpose, emissions data based on air–fuel ratio (AFR), engine speed (RPM) and different fuel flow rate parameters are used experimentally under different loads. In the designed LSTM models, fuel flow, engine speed and AFR are used as input parameters for the prediction of exhaust emission indices, engine speed and AFR data is used as an input parameter for the prediction of combustion efficiency. In the designed system, the experimental data is divided into two, 80% training and 20% test, by crossing according to the k-fold 5 value. Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) error functions and the R squared (R 2 ) function are used in the evaluation of the designed LSTM models. The originality of this study is the prediction of the exhaust emissions index and combustion efficiency values for the T56-A-15 turboprop engine using the LSTM method, which is an artificial intelligence method. This study attempts to address the literature gap in the calculation of the CO, CO2, UHC, NO2 exhaust emissions index and combustion efficiency values with an accuracy of over 95%, without the need for hundreds of experimental studies required for intermediate values.