基于CEEMDAN-MFDE-HHO-SVM的机载燃油泵故障辨识

Fault identification of airborne fuel pump based on CEEMDAN-MFDE-HHO-SVM

针对机载燃油泵振动信号的有效分量相互耦合、故障特征提取困难,进而导致故障识别准确率低的问题,提出了一种基于自适应噪声完备经验模态分解(CEEMDAN)、多尺度波动散布熵(MFDE)和哈里斯鹰算法(HHO)优化支持向量机(SVM)的机载燃油泵故障辨识方法(CEEMDAN-MFDE-HHO-SVM)。首先,采用CEEMDAN方法对机载燃油泵振动信号进行了自适应分解,生成了一组从低频到高频分布的本征模态函数(IMF),并选择包含冲击信息较多的IMF分量进行了信号重构,得到了噪声含量较低的信号;然后,采用MFDE方法计算了低噪信号的熵值,构造了表征样本故障属性的特征矩阵;最后,采用HHO算法对SVM的关键参数进行了优化,以构造基于HHO-SVM模型的多故障分类器,对机载燃油泵的故障进行了辨识;基于实测机载燃油泵故障数据集,将CEEMDAN-MFDE-HHO-SVM方法与其他组合方法进行了对比分析。研究结果表明:该故障辨识模型的故障分类准确率达到了100%,在信号处理、熵值特征提取和分类器方面都优于其他对比方法;该模型不仅具有更高的分类准确率,而且具有更优异的效率,后续可以将其推广到其他机械设备的故障辨识中。

Aiming at the problem that effective components of vibration signal of airborne fuel pump were coupled to each other and fault feature extraction was difficult,which led to low fault identification accuracy,a fault identification method based on complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN),multiscale fluctuation dispersion entropy(MFDE),and Harris hawk algorithm(HHO)optimized support vector machine(SVM)was proposed(CEEMDAN-MFDE-HHO-SVM).Firstly,CEEMDAN method was used to decompose the airborne fuel pump vibration signal,generating a set of intrinsic mode functions(IMF)distributed from low frequency to high frequency,and IMF components containing more impact information were selected for signal reconstruction to obtain signals with lower noise content.Then,the MFDE method was used to calculate the entropy value of the low noise signal and a feature matrix that characterized the fault attributes of the sample was constructed.Finally,the HHO algorithm was used to optimize the key parameters of SVM to construct a multi-fault classifier based on HHO-SVM model,and fault identification of the airborne fuel pump was completed.A comparative analysis was conducted between the CEEMDAN-MFDE-HHO-SVM method and other combination methods based on the measured airborne fuel pump fault dataset.The results show that the classification accuracy of the fault identification model reaches 100%,which is superior to other comparison method in signal processing,entropy feature extraction and classifier.It not only has higher classification accuracy but also has better efficiency,which can be extended to other mechanical equipment fault identification in the future.