基于CEEMD香农熵和GAPSO-SVM的机载燃油泵故障诊断方法

FAULT DIAGNOSIS METHOD OF AIRBORNE FUEL PUMP BASED ON CEEMD SHANNON ENTROPY AND GAPSO-SVM

机载燃油泵是燃油系统的关键组成部件,针对信号分解重构过程中出现的模态混叠和残余分量过大现象,提出了一种基于CEEMD香农熵和改进SVM的机载燃油泵故障诊断方法。通过机载燃油泵故障诊断实验台,获取其各种工况下的壳体振动和出口压力信号,使用CEEMD方法进行振动信号分解,计算所得IMF的香农熵,以此为依据,选取合适的信号提取能量值和压力信号均值作为SVM的输入特征向量,采用遗传粒子群算法(GAPSO)优化的SVM诊断燃油泵的故障类型。将结果与BP神经网络、粒子群算法优化的SVM、遗传算法优化的SVM等方法进行比较,结果表明遗传粒子群算法优化的SVM诊断方法模型训练快、准确率高、时效短,具有良好的工程应用价值。

The airborne fuel pump is a key component of the fuel system.In view of the phenomena of mode aliasing and excessive residual component in the process of signal decomposition and reconstruction,a fault diagnosis method for airborne fuel pump based on CEEMD Shannon entropy and improved SVM is proposed.The signals of shell vibration and outlet pressure under various working conditions are obtained on the fault diagnosis test bench of airborne fuel pump.Then in the simulation I decomposed the vibration signals by using CEEMD method and calculated the Shannon entropy of IMF.Based on the above results,I selected the energy value and the mean value of pressure signal as the input eigenvectors of SVM,and used the SVM optimized by Gapso the to diagnose the fault types of fuel pump.Compared with BP neural network,SVM optimized by particle swarm optimization(PSO)and SVM optimized by genetic algorithm(GA),the results showed that the model of SVM diagnosis optimized by GA has the advantages of fast training,high accuracy and short time-effect,and it has good engineering application value.