The cone-shaped kernel distributions o f vibration acceleration signals, which were acquired from the cylinder head in ei ght different states of a valve train, were calculated and displayed in grey ima ges. Probabilistic Neural Networks (PNN) was used to classify the images directl y after the images were normalized. By this way, the problem of fault diagnosis for a valve train was transferred to the classification of time-frequency image s. As there is no need to extract features from time-frequency images before cl assification, the fault diagnosis process is highly simplified. The experimental results show that the vibration signals can be classified accurately by the pro posed methods.
展开▼
