论文部分内容阅读
针对传统的液体火箭发动机涡轮泵故障诊断方法只能在有样本数据并且样本数据充足的情况下才能进行准确诊断以及诊断时难以提取状态特征的缺点,提出一种适用于涡轮泵在线监测及诊断方法,该方法利用生物免疫系统的反面选择机理,利用生物克隆和学习机理使改进型反面选择算法产生的检测器具有不同的检测半径,使其能更有效地覆盖异常空间,能有效地提取涡轮泵的状态特征,避免了检测器产生效率低等问题。实例诊断结果表明:该方法较好地解决了故障样本难以获取及有效地提取涡轮泵的状态特征的问题,能准确监测出涡轮泵各种常见故障所引起的异常并能准确诊断,较高诊断精度表明该方法是可行的,并且具有较好的在线性、准确性及鲁棒性,为液体火箭发动机涡轮泵故障异常检测探索了一条新路。
Aiming at the shortcomings that the traditional liquid rocket engine turbo-pump fault diagnosis method can only perform accurate diagnosis with the sample data and sufficient sample data, and it is difficult to extract the state features at the time of diagnosis, a method suitable for on-line monitoring and diagnosis of turbo-pumps This method makes use of the negative selection mechanism of biological immune system and utilizes the biological clone and learning mechanism to make the detectors produced by the improved reverse selection algorithm have different detection radiuses so that it can more effectively cover the abnormal space and effectively extract the turbo pump State characteristics, to avoid the detector produces low efficiency and other issues. The experimental results show that this method can solve the problem of difficult to obtain the fault samples and effectively extract the characteristics of the turbo pump. It can accurately detect the turmoil caused by various common faults and can accurately diagnose the turbo pump. Accuracy shows that this method is feasible, and has good linearity, accuracy and robustness, and explores a new approach for fault detection of turbopump in liquid rocket engine.