基于ADEGWO-SVM的机载燃油泵寿命预测研究

Research on remaining useful life prediction of fuel pump based on adaptive differential evaluation grey wolf optimizer-support vector machine

针对高可靠、长寿命、小样本的机载燃油泵剩余寿命预测问题,提出了一种基于自适应差分变异的狼群支持向量机（ADEGWO-SVM）组合寿命预测方法。首先,搭建了一个机载燃油泵寿命试验平台,监测其电应力载荷下的出口压力信号,应用小波包滤波的方法对原始压力信号进行降噪处理,提取压力均值信号作为燃油泵性能退化特征,接着利用自相关分析的方法进行特征相空间重构;然后利用基于ADEGWO算法结构简单和全局搜索能力的特点,优化支持向量机预测模型的参数,进而提出一种基于ADEGWO-SVM的组合寿命预测模型;最后,在不同的预测起始点,利用ADEGWO-SVM方法进行了寿命预测试验,为了进一步验证该算法的有效性,将其与粒子群支持向量机（PSO-SVM）,灰色模型（GM（1,1））等算法进行了比较,试验结果表明,该方法能够准确实现机载燃油泵的剩余寿命预测,显著提高寿命预测精度,对机载燃油泵的健康监测和寿命预测具有理论指导意义。

Aiming to solve the problem of predicting the remaining useful life（ RUL） of high-reliability,long-life,and small-sample airborne fuel pump,an improved RUL prediction method based on adaptive differential evaluation grey wolf optimization-support vector machine（ ADEGWO-SVM） is proposed. Firstly,an airborne fuel pump life test platform is set up to monitor the outlet pressure signal under the electrical stress load. The wavelet packet filtering method is used to denoise the original pressure signal. The average pressure signal is extracted as the degradation feature of the fuel pump. Then,the feature phase space reconstruction is performed by the adaptive correlation analysis. By using the characteristics of the simple structure and global search ability of the GWO algorithm based on adaptive differential evaluation,the parameters of SVM prediction model are optimized. The combined RUL prediction model based on ADEGWOSVM is proposed. Finally,the ADEGWO-SVM method is adopted to predict the life prediction at different prediction starting points. To further evaluate the algorithm,it is compared with the PSO-SVM,GM（ 1,1）,and other algorithms. Experimental results show that ADEGWO-SVM method can accurately predict RUL of the airborne fuel pump,significantly improve the prediction accuracy,which has theoretical significance for the health monitoring and life prediction of the airborne fuel pump.