液体火箭发动机高速涡轮泵的振动故障检测

讨论了涡轮泵故障的几个主要原因，据此提取涡轮泵振动数据的特征，用BP神经网络的方法进行故障检测。BP神经网络的训练样本集由一个具有无监督聚类功能的神经网络从原始的特征向量集获取。

FAULT DETECTION OF HIGH SPEED TURBOPUMP VIBRATION IN LIQUID PROPELLANT ROCKET ENGINE

Some main causes of occurred faults in the fuel trubopump of a rocket engine arediscussed. According to these faults, some features for fault detection are summarized from the powerspectrum density of a channel of vibration acceleration signal measured from the turbopump case.Neural networks are used to detect faults. With a BP neural network acting as fault detector, anunsupervised clustering neural network is used to obtain patterns required for training the faultdetector from the original test data characteristic vectors. The validation of the vibration signal inengine ground firing test shows the feasibility and effectiveness of this method.