Knock Intensity Diagnosis by Vibration Analysis Using Ensemble Empirical Mode Decomposition

Knock is an undesirable event that causes most of the abnormal combustion in spark ignition (SI) engines. In this paper, a novel knock intensity diagnosis method for SI engines by vibration analysis based on the ensemble empirical mode decomposition (EEMD) is proposed. First, the vibration signal measured from the test SI engine cylinder head is decomposed by the EEMD. Because the EEMD can eliminate the mode mixing problem existing in the classic empirical mode decomposition (EMD), the EEMD method can provide an improved time-scale decomposition result with a clearly physical meaning for the individual IMFs. Second, a series of temporal and frequency statistical characteristics are calculated from the IMF component in which the knock characteristics appear including peak value, standard deviation, shape factor, kurtosis, crest factor, etc. Finally, the obtained SI engine knock characteristics vectors are input to the classifiers to accomplish knock intensity identification. Experimental results indicate that the proposed method can achieve high knock intensity diagnosis accuracy even under the condition of a very high engine speed of 5800 rpm.