Prediction of combustion pressure for a dual-cylinder free-piston engine generator based on data driven

The free-piston engine generator (FPEG) is an advanced linear power device renowned for its high energy conversion efficiency. In a dual-cylinder FPEG, the peak combustion pressure near the top dead center determines the piston's maximum acceleration, significantly impacting the engine's expansion stroke. However, the absence of mechanical constraints results in unstable FPEG operation and considerable variations in the combustion cycle, making it difficult to measure or predict the uncertain peak combustion pressures in advance. This study proposes a data-driven method to predict peak combustion pressure for early diagnosis and exception warning. Historical in-cylinder pressure data from a dual-cylinder FPEG prototype are analyzed to extract two key features: slope and interval features. These features show an average linear correlation degree above 0.98, as indicated by the Pearson coefficient. Statistical analysis is used to create regression models based on these features. By combining regression features with the ignition pressure, a prediction model for peak combustion pressure is derived. Tests under various operating conditions demonstrate the model's accuracy and robustness, with average prediction error indices (MAE, MSE, RMSE) consistently below 0.05.