Misfire Detection in Automotive Engines Using Structure-Borne Sound

Though a variety of misfire detection methods exists, the number of approaches based on structure-borne sound is scarce due to the difficult interpretation of the accelerometer signals. The procedure presented in this paper combines the EM (Estimation-Maximization) algorithm and a simple parametric pressure model to decompose the structure-borne sound into the individual contributions of each combustion. This decomposition is exploited to develop a simple statistical test for misfire detection based on the shape of the reconstructed pressure. Though it has to be adjusted slightly with empirical data for optimal performance, the test automatically yields an identification threshold as a function of the engine speed and the desired detection probability. Measurement data collected from a four cylinder, 1.8 l Audi SI test bed engine are used to validate the proposed approach.