| 摘要: |
| 针对运载火箭动力系统故障的复杂特性和故障模式的不确定性，提出了基于神经网络和证据理论的火箭发动机故障诊断。首先以火箭视加速度和角速度作为网络输入，故障类型矩阵作为网络输出，通过BP神经网络和RBF（径向基函数）神经网络进行故障诊断；之后通过D-S证据理论融合神经网络结果；最后通过滚动时域估计方法对火箭飞行状态特征量估计。仿真结果表明诊断准确率达到99%以上，表明提出的方法对于火箭发动机故障诊断具有较高的准确性和实用性。 |
| 关键词: 火箭发动机 神经网络 证据理论 滚动时域估计 故障诊断 |
| DOI： |
|
| 基金项目: |
|
| Fault Diagnosis of Rocket Engine Based on Neural Network and Evidence Theory |
| SUN Chengzhi,YAN Xiaodong |
| (School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China;School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China; Shanxi Aerospace Flight Vehicle Design Key Laboratory, Xi'an 710072, China) |
| Abstract: |
| Due to the complexity of the failure situations of the launch vehicle's propulsion system, the failure is usually hard to diagnose onboard accurately. A diagnose method for the failure of the rocket engine is proposed based on neural network and evidence theory in this paper. Firstly, the BP neural network and the RBF neural network are trained offline. The apparent accelerations and angular velocities of the dynamics are taken as the network input, and the fault type matrix is taken as the network output. Then, the BP neural network and the RBF neural network are used to implement the fault diagnosis online. Then the D-S evidence theory are used to fuse the neural network results. Finally, the fault characteristic parameters are estimated by the moving horizon estimation algorithm. The simulation results show that the diagnostic accuracy is over 99%, which indicates that the proposed method has higher accuracy and practicability for rocket engine fault diagnosis. |
| Key words: Rocket engine Neural network Evidence theory Moving horizon estimation Fault diagnosis |