基于改进IRN网络的航空发动机故障诊断

Aeroengine Fault Diagnosis Based on Improved Internal Recurrent Neural Networks

建立了一种带偏差单元的内部回归神经网络。该网络的主要特点是记忆特性好 ,收敛速度快、稳定性强。介绍了该神经网络在航空发动机故障诊断中的应用方法 ,并通过实例验证了该方法在发动机故障分类中的实用性。在对某型航空发动机故障诊断中 ,准确性极高。从而在适时与视情维修过程中提出合理的建议 ,以便为相应维修方法的实施、改进和研究应用提供可靠的依据 ,有着良好的推广应用前景。

A novel internal recurrent neural network with partial cell is presented in this paper. This network possesses virtues of good memory,fast convergence ability and strongly stable capability. The presented networks are introduced to aeroengine fault diagnosis;their application testifies this method is practical and instructional. Results show the method owns good robustness and high precision. This method is instructional and contributive to maintenance and logistics.