Abstract: In order to extract the diesel engine fault characteristics from the non-stationary and nonlinear cylinder head vibration signal, a fault diagnosis method for diesel engine misfire based on VMD and multi-scale dispersion entropy is proposed. The vibration signal of diesel cylinder head is decomposed by VMD, and then the modal component with the minimum dispersion entropy is selected as the analysis signal to calculate the multi-scale dispersion entropy (MDE). The first six scales of dispersion entropy as fault eigenvector are input into the support vector machine of particle swarm optimization (PSO-SVM ) for misfire fault classification and judgment, which is compared with other four common methods. The results show that the proposed diagnostic method can effectively extract fault features and accurately identify fault types, which is better than the methods compared.

Key words: diesel engine, fault diagnosis, VMD, MDE, PSO-SVM