A Method of Adaptive Neuron Model (AUILS) and Its Application
Zhai Jun, Xiaojia Yang, Yan Chen · 2006
This paper presents an adaptive neuron model utilizing information of local samples - the AUILS neuron model. Differing from traditional neuron models, the AUILS neuron model fully employs the experience samples information within the local range and well embodies the association and analogy functions of cerebrum. The neural network, which is simple in structure and fast in learning speed, can realize the nonlinear mapping relationship between multi-input and multi-output. Through investigating the properties of AUILS and learning algorithm based on gradient, we build a method based on the neuron model for rotary machine fault diagnosis, which takes full advantage of expert experiences to estimate the reliability of fault existing according to the vibration signals of rotary machine in operation. It is verified that using the AUILS model, expert experiences can be well expressed in the diagnosis results