Study of the model of multi-layer fusion diagnosis based on neural network and fuzzy integral

Aiming at the weakness of the fusion diagnosis model based on neural network, this paper provides a model of multi-layer fusion diagnosis based on neural network and fuzzy integral. This model consists of two layers: 1. The first layer is local fusion diagnosis layer that adopts the information fusion technology based on neural network to realize the local information diagnosis. 2. The second layer is global fusion diagnosis layer based on fuzzy integral. The inputs are the local decisions from subnet. The fuzzy integral is used to realize the final global diagnosis and the diagnosis conclusion is gotten in the end. In this paper, the model of multi-layer fusion diagnosis based on neural network and fuzzy integral is used in the fault diagnosis of rolling bearing. The experiment result shows that in comparison with present multi-sensor fusion diagnosis model based on neural network, this model has made great improvements in such aspects as constructing and training neural network as well as the real-time quality, robustness of the diagnosis system.