Fault diagnosis for diesel valve trains based on non-negative matrix factorization and neural network ensemble

It is well known that the vibration signals are unstable when there is some failure in machinery. So in this paper, the cone-shaped kernel distributions (CKD) of vibration acceleration signals acquired from the cylinder head in eight different states of valve train were calculated and displayed in grey images. Meanwhile, non-negative matrix factorization (NMF) was used to decompose multivariate data, and neural network ensemble (NNE), which is of better generalization capability for classification than a single neural network, was used to perform intelligent diagnosis without further fault feature (such as eigenvalues or symptom parameters) extraction from time–frequency distributions. It is shown by the experimental results that the faults of diesel valve trains can be accurately classified by the proposed method.