基于模糊神经网络的液体火箭发动机振动检测

液体火箭发动机振动检测涉及部件振动数据的收集、振动特征的抽取与度量以及度量结果的决策。基于模糊神经网络提出了一种发动机振动故障检测的基本系统。这种技术的吸引力在于：神经网络采用可变模糊集代表发动机工作模式，自然地提供了反映故障程度的有用信息；神经网络的离线学习算法可以从训练样本中提取振动知识；神经网络的监测算法不仅能正确预报故障，同时也能对新的振动信息进行在线学习。实验研究结果表明：模糊神经网络可以成功地用于泵压式液体火箭发动机热试车的振动故障检测。

Vibration Detection of Liquid Rocket Engine Based on Fuzzy Neural Networks

The vibration detection of a liquid rocket engine involves the collection of vibration data from the engine components,the extraction and measurement of vibration features that reflect the operational state of rocket engine and the decision making of measuring results.Based on fuzzy neural networks,a basic system used for vibration fault detection of rocket engines is proposed in this paper.This technology provides an attractive implementation to vibration detection because of the following.Neural networks utilize fuzzy sets as operational modes fo rocket engine so that they inherently provide useful information that represents for engine fault degree.The engine vibration knowledge is automatically elicited from the training pattern sets in the off line learning process of neural networks.The monitoring algorithm of neural networks not only correctly gives an alarm for indicating the existence of the engine faults but also can learn the new vibration messages from the input data for the on line conditions.The results of experimental research have shown that the fuzzy neural networks can be successfully employed in the vibration fault detection in the firing tests of a turbo pump liquid rocket engine.