Misfire and knock detection based on the ion current inside a passive pre-chamber of gasoline engine

Abstract

In this study, a novel idea of capturing ion current signals inside the passive pre-chamber is proposed. The signals were analyzed in terms of various operating conditions based on a naturally aspired gasoline engine. Under normal combustion, the chemical and the thermal ionization peak's phase difference is much smaller than the commercial spark plug case. Under the idle condition, misfire cycles by cutting off fuel supply can be detected by integrating ion current signal, with 100% diagnosing sensitivity and its accuracy reaching 99.7%. The candidate knocking cycle is defined as a combustion cycle with an in-cylinder pressure rising rate higher than 0.2 MPa/°CA. Under candidate knocking cycles, a unique knock indication peak from the ion current signal can be noticed. To classify candidate knocking cycles from normal combustion, the phase difference of chemical and the last ionization peak can be regarded as a basis with the sensitivity of 99.7% and the accuracy of 95.7%. For further increasing the diagnostic accuracy, several artificial neural networks were established utilizing various ion current parameters. Through 5-fold cross-validation, with the best-performed combination of the ion current parameters as the input variables, the diagnosis model further improves the diagnostic accuracy from 95.7% to 98.4% with the same sensitivity of 99.7%. Utilizing 3-dimensional simulation, the cause of the knock indication peak in the ion current signal can be explained by the visualized backflow from the main chamber and the traces of critical ions and electrons concentration inside the pre-chamber. The employment of the ion current inside the passive pre-chamber for abnormal combustion detection thus appears very promising.