基于WOA-VMD-MSE-SVM的海水泵激励源识别方法  

Seawater pump excitation source identification based on WOA-VMD-MSE-SVM

作　　者：滕佳篷 武国启 富琦晋 TENG Jiapeng;WU Guoqi;FU Qijin(Dalian Measurement and Control Technology Research Institute,Dalian 116013,China)

机构地区：[1]大连测控技术研究所,辽宁大连116013

出　　处：《舰船科学技术》2024年第18期44-48,共5页Ship Science and Technology

基　　金：基础产品创新科研资助项目(0207024)。

摘　　要：针对海水泵的复杂多激励源难以准确识别问题,提出一种基于变分模态分解(VMD)、鲸鱼优化算法(WOA)、多尺度熵(MSE)和支持向量机(SVM)相结合的海水泵激励源识别方法。基于海水泵振动信号,首先采用VMD和WOA算法获取分解层数K与惩罚系数α两个重要参数,并对信号进行分解重组;然后提取重组信号的多尺度熵,作为WOA-SVM模型的输入特征向量,并对SVM的惩罚因子c和核系数g两个重要参数进行寻优,最后将得到的参数代入SVM模型进行训练与激励源识别。通过实船海水泵激励源识别及对比分析,验证识别方法的有效性。研究结果表明提出的WOA-VMD-MSE-SVM算法满足海水泵激励源识别准确要求。Aiming at the problem that it is difficult to identify the complex multiple excitations of the sea water pump accurately,a method based on the combination of variational modal decomposition(VMD),whale optimisation algorithm(WOA),multi-scale entropy(MSE)and support vector machine(SVM)is proposed to identify the excitation source of the sea water pump.Based on the vibration signal of the sea water pump,the WOA algorithm and VMD algorithm are firstly used to optimise the two important parameters,the number of decomposition layers K and the penalty coefficientα,then the signal is decomposed and reorganised.The multiscale entropy of the restructured signal is extracted and input into the WOA-SVM model as a feature vector to perform parameter optimisation on the two important parameters of the SVM,the penalty factor c and the kernel coefficient g.Finally,the obtained coefficients are brought into the SVM model for training and excitation source identification.Through the real ship seawater pump excitation source identification and comparative analysis,verified that the identification method is effective.The research results surface that the proposed WOA-VMD-MSE-SVM algorithm meets the accurate requirements for the identification of excitation sources of seawater pumps.

关 键 词：激励源识别 海水泵 支持向量机 多尺度熵 鲸鱼优化算法 变分模态分解 

分 类 号：U661.44[交通运输工程—船舶及航道工程] TH165.3[交通运输工程—船舶与海洋工程]