Experimental study of the combustion characteristics prediction model for a sensor-less closed-loop control in a heavy-duty NG engine

Abstract The closed-loop combustion control is an effective way to reduce the cycle-to-cycle variations of spark ignition (SI) engine. In this paper, a series of experimental tests were conducted to investigate the combustion performance and emissions characteristics in a six-cylinder natural gas (NG) engine. The correlation between combustion phasing and control parameters was analyzed under different EGR rates and spark timing (ST) at medium and low loads. Additionally, a prediction model on the basis of the correlation was developed employing artificial neural network (ANN) for a sensor-less closed-loop combustion control. After analyzing the results, it is found that a linear relationship exists between CA 50 and ST under various EGR rates. Engine out emissions (i.e., CO, THC, and CH4) are insensitively affected by the CA50. It is observed that the NOx emission is significantly higher in comparison to those under the higher EGR rate. As a consequence, CA50 is acting as the feedback signal for closed-loop combustion control with consideration of cylinder consistency. Moreover, the ANN model was proven to be an impressive way to predict the combustion characteristics, which can substitute the cylinder pressure sensor. Moreover, the prediction model can be applied in a sensor-less closed-loop combustion control, while the correlation coefficient value can reach up to approximately 1.00.