基于小波包BP_AdaBoost算法的机载燃油泵故障诊断研究

针对机载燃油泵故障数据少、诊断效率低、维护成本高、缺乏有效诊断方法的问题,搭建了机载燃油泵燃油转输系统实验平台,提出利用小波包分析进行特征提取和基于BP_AdaBoost机载燃油泵故障诊断方法。首先测量燃油泵7种典型状态模式所对应的振动信号和出口压力信号;然后在分析信号时频特性和统计特性的基础上,利用小波包分解提取振动信号不同频段能量值作为故障特征参数,结合振动信号峭度以及压力信号均值构造特征向量;最后利用特征向量训练和验证BP_AdaBoost分类模型。实验结果不仅优化了传感器,而且表明BP_Adaboost算法与SVM、BP算法相比,能够有效实现对机载燃油泵的故障诊断。

Research on fault diagnosis for airborne fuel pump based on wavelet package and BP_AdaBoost algorithm

Aiming at the problems of less failure data, low diagnostic efficiency, high maintenance cost and lack of efficient diagnosis method of airborne fuel pump, an experiment platform of the fuel transfer system for airborne fuel pump is developed and a fault diagnosis method for airborne fuel pump based on wavelet packet analysis and BP_AdaBoost neural network algorithm is presented. Firstly, the vibration signals and outlet pressure signals of the fuel pump corresponding to seven kinds of typical state modes are acquired. Then, on the basis of the analyzing of the signal time frequency features and statistical features, wavelet packet decomposition is used to extract the energies of the vibration signal in different frequency bands that are taken as the fault characteristic parameters, which combines with the vibration signal kurtosis and mean outlet pressure to construct the fault feature vectors. Finally, the fault feature vectors are used to train and verify the BP_AdaBoost classification model. The experiment results not only optimize the sensor, but also show that the BP_AdaBoost algorithm can effectively achieve the fault diagnosis of airborne fuel pump compared with SVM and BP algorithms.