A Study on Predicting Engine Performance Outputs by Machine Learning Algorithms in a Single Cylinder HCCI Engine

Machine learning algorithms are often used to mathematically establish relationships between data sets. Successful results have been achieved in performance, production, consumption, fault, and wear prediction applications using learning algorithms. The high testing costs of Homogeneous Charge Compression Ignition (HCCI) engines, the determination of efficient operating ranges, and the challenges of performance prediction in untested regions have recently made the use of artificial intelligence technologies increasingly popular. In this study, a dataset (805 data) was created by varying the λ value in a single-cylinder HCCI engine (Ricardo Hydra) and conducting performance measurements at different engine speeds. Based on the input values of Compression Ratio, RON (Research Octane Number), Intake Air Temperature (K), Engine Speed (rpm), and Lambda (λ) within the dataset, the output variables IMEP(Bar), Effective Torque, Indicated Thermal Efficiency, and COVimep (%) were predicted. In this study, a prediction model was developed using the AdaBoost and Tree machine learning algorithms. The experimental results demonstrated that the AdaBoost algorithm achieved the highest accuracy in predicting IMEP (Bar) output values and the lowest error rates in predicting Indicated Thermal Efficiency output values. The highest performance was obtained with an R metric a value of 9.57×10-1, while the lowest error rates were calculated as 2.89×10-4 for the MSE error metric, 1.70×10-2 for the RMSE error metric, 1.30×10-2 for the MAE error metric, and 4.10×10-2 for the MAPE error metric. The results indicate that high-accuracy predictions can be made using the proposed model.