基于MFCC与CNN的机械故障声音自动识别  

Automatic Recognition of Mechanical Fault Sound Based on MFCC and CNN

作　　者：黄炜 罗谢飞 HUANG Wei;LUO Xiefei(Guangxi Business&Trade Technician College,Nanning 530000,China)

机构地区：[1]广西商贸技师学院,广西南宁530000

出　　处：《电声技术》2024年第6期129-131,共3页Audio Engineering

摘　　要：针对机械故障自动识别问题,提出一种结合梅尔频率倒谱系数(Mel Frequency Cepstral Coefficient,MFCC)与一维卷积神经网络(Convolutional Neural Networks,CNN)的机械故障声音自动识别方法,并通过实验验证该方法的有效性。实验结果表明,该方法在机械故障声音识别中具有较高的准确率、精确率及召回率,能够有效识别故障案例。This paper proposes a mechanical fault sound automatic recognition method that cmbines Mel Frequency Cepstral Coefficient(MFCC)and one-dimensional Convolutional Neural Networks(CNN)for the problem of mechanical fault automatic recognition.The effectiveness of this method is verified through experiments.The experimental results show that this method has high accuracy,precision,and recall in mechanical fault sound recognition,and can effectively identify fault cases.

关 键 词：机械故障 声音识别 梅尔频率倒谱系数(MFCC) 卷积神经网络(CNN) 

分 类 号：TH17[机械工程—机械制造及自动化]