ACTA AERONAUTICAET ASTRONAUTICA SINICA ›› 2016, Vol. 37 ›› Issue (9): 2851-2863.

• Material Engineering and Mechanical Manufacturing •

### Fault states feature extraction and experimental study for airborne fuel pumps based on sample quantile

LI Juan1,2, JING Bo1, QIANG Xiaoqing1, LIU Xiaodong3,4

1. 1. College of Aeronautics and Astronautics Engineering, Air Force Engineering University, Xi'an 710038, China;
2. College of Mathematics and Statistics, Ludong University, Yantai 264025, China;
3. Nanjing Engineering Institute of Aircraft Systems, Jincheng, AVIC, Nanjing 210000, China;
4. Aviation Science and Technology Key Laboratory of Aviation Mechanical and Electrical System, Nanjing 210000, China
• Received:2015-08-31 Revised:2015-11-12 Online:2016-09-15 Published:2015-11-27
• Supported by:

Aeronautical Science Foundation of China (20142896022)

Abstract:

The health state of airborne fuel pumps is important for the achievement and safety of flight mission, so the fault state feature extraction and diagnosis for the fuel pumps become an urgent problem. Through the analysis of the output signal about the vibration sensors and pressure sensor from the experiment system of airborne fuel pump, a sample quantile based fault state feature extraction method is presented. Firstly, by the asymptotic distribution theorem of the quantile, the statistic character of the sample quantile is discussed; the correspondence between the fault state and the sample quantile is discussed; the feasibility of the presented method is ensured in theory; statistic analysis has been done for the real data, and the stability alone with the sample size is discussed. Secondly, the confidence interval of sample quantile is computed based on the asymptotic distribution theorem, and compared with that obtained by the Bootstrap methods. The results show that the confidence interval of sample quantile obtained according to the asymptotic distribution theorem is credible, providing basis for online fault diagnosis. Then the Bayesian discriminant function is established based on the quantile feature vector extracted under different states, which will be used for the fault diagnosis. Finally, the layout of the sensors is optimized based on the accuracy of fault diagnosis. The results show that the vibration sensors combined with pressure sensor will improve the accuracy of fault diagnosis, and only one vibration sensor combined with one pressure sensor can finish the fault diagnosis completely. This study provides theoretical support for the quick and accurate online decision about the states of the fuel pump, and it can reduce volume, power consumption and complexity of the system in engineering applications.

CLC Number: