基于SABO-VMD-Adaboost-RF的直流电机噪声源识别  

Identification of DC Motor Noise Sources Based on SABO-VMD-Adaboost-RF

作　　者：祁丽春 QI Lichun(Changshu Branch of Jiangsu United Vocational and Technical College,Changshu 215500,China)

机构地区：[1]江苏联合职业技术学院常熟分院,江苏常熟215500

出　　处：《常熟理工学院学报》2025年第2期75-83,共9页Journal of Changshu Institute of Technology

摘　　要：原始噪声信号中包含了多种电机噪声激励源以及环境噪声,使得从信号中识别出各个噪声源变得困难.为解决这一问题,本研究提出了一种基于SABO(Subtraction-Average-Based Optimizer,SABO)优化算法的变分模态分解(Variational Mode Decomposition,VMD)结合经过增强学习(Adaboost)迭代优化的随机森林(Random Forest,RF)的方法.首先,通过SABO优化算法寻找VMD的最佳参数并代入VMD算法中,利用优化后的VMD对原始信号进行处理.然后对数据进行特征提取,并利用Adaboost算法对随机森林RF进行迭代优化,最后利用优化后的RF对这些特征数据进行训练和分类识别.结果表明,该方法能够准确地识别出电机的正常噪声信号、由转子不平衡引起的机械振动噪声过大的信号、由径向电磁力波引起的噪声过大的信号和由轴承装配不稳产生的噪声信号,为直流电机降噪及结构优化提供了理论依据.During the noise testing process of DC motors,the raw noise signals contain various motor noise excitation sources as well as environmental noise,making it difficult to identify each noise source from the signals.To address this issue,this study proposes a method that combines Variational Mode Decomposition(VMD)optimized by the SABO algorithm with a Random Forest(RF)that has undergone iterative optimization through enhanced learning(Adaboost).Firstly,the optimal parameters for VMD are found by using the SABO optimization algorithm and then by applying it to the VMD algorithm to process the raw signals.Subsequently,feature extraction is performed on the data,and the RF is iteratively optimized by using the Adaboost algorithm.Finally,the optimized RF is used to train and classify the feature data.Experimental results show that this method can accurately identify the normal noise signals of the motor,signals with excessive mechanical vibration noise caused by rotor imbalance,signals with excessive noise caused by radial electromagnetic force waves,and noise signals produced by unstable bearing assembly.The accuracy rate reaches 99.57%.This noise source identification technology based on SABO-VMD-Adaboost-RF provides a theoretical basis for noise reduction and structural optimization of DC motors.

关 键 词：直流电机 噪声源识别 SABO优化算法 变分模态分解 增强学习 随机森林 

分 类 号：TN820.1[电子电信—信息与通信工程]