带有偏差单元的递归神经网络在故障诊断方面的应用

Application of Nerve Network Model with Deviation Unit for Fault Diagnosis

针对多层BP网络的不足，在BP网的基础上，给出一种带有偏差单元的递归神经网络模型及其算法，在学习过程中引入经验知识。仿真表明该算法能大大提高学习速度和训练精度。

Aim at the shortages of multi-layer BP network and based on the BP network, this paper provides a recursive nerve network model with deviation unit and its algorithm, which introduces the experiential knowledge in the learning process. The emulation demonstrates that this algorithm can definitely enhance the learning speed and training accuracy.