Detection of Misfire through Knock sensor signal.

Constraints imposed in the last decades on emission reductions heavily influenced the evolution of technologies whose work rests on internal combustion engines. Stricter carbon dioxide emission regulations shifted the focus towards fuels like Hydrogen, with the consequent rise of uncertainties regarding the combustion profile. One of the crucial points analyzed by INNIO, an Austrian company manufacturing power-generating units, was the instability of a fuel like Hydrogen and the possible risk of compromising the lifespan of essential ICE components. Given the imprevedibilities and abnormalities encountered during the combustion process, a way to identify missing combustion cycles, called misfires is a crucial topic. Through the study of previous works, signal analysis, and continuous wavelet transformation methods, it is possible to obtain noteworthy results following different manipulations of the raw data generated by a piezoelectric knock sensor installed on the cylinder head of an ICE. Upon completion and validation of a theoretical approach identifying combustion through the use of recurring patterns over different working conditions and typologies of engines, it would be possible to detect misfires, act against fuel waste, reduce emissions, and prevent component failures. Combining an algorithm with machine learning procedures and regression models, such as XGBoost, a reduction in the use of expensive sensors to monitor the in-cylinder pressure will improve adaptations to cylinder-cylinder variations in multi-cylinder engines.