Pressure reconstruction and misfire detection from multichannel structure-borne sound

Though a variety of misfire detection methods exist, the number of approaches based on structure-borne sound is scarce due to its difficult interpretation. The paper presents a procedure for misfire diagnosis that exploits a multichannel sound signal from accelerometers mounted on the surface of an automotive engine. The EM algorithm is used to decompose the measured sound into individual components associated with single combustions. For these simplified scenarios, a statistical test based on the shape of the reconstructed in-cylinder pressure is devised. Measurement data collected from a four cylinder test bed engine is used to validate the proposed approach and show its superiority to the single-channel solution.