燕山大学学报2025，Vol.49Issue(5)：404-413,10.DOI:10.3969/j.issn.1007-791X.2025.05.003
基于MFCC-LSTM的低速齿轮故障诊断方法研究
Research on low-speed gear fault diagnosis method based on MFCC-LSTM
摘要
Abstract
To address the challenges of acquiring subtle fault signal features and low fault diagnosis accuracy in low-speed gear operation,a method for diagnosing low-speed gear faults is proposed,which combines Mel-frequency cepstrum coefficient(MFCC)with long short-term memory(LSTM)network.Firstly,the low-frequency energy of the input signal is extracted by MFCC,and then the feature vectors are input into the LSTM network for training to obtain the MFCC-LSTM model,which plays the advantages of MFCC for low-frequency energy extraction and LSTM network for long sequence data processing,and enhances the accuracy of low-speed gear fault diagnosis.Finally,the performance of the proposed method is verified by acquiring acoustic sensing signals of low-speed gear operation through the rotating machinery fault simulation experimental platform,and the optimal number of Mel filter groups of MFCC is obtained from the experiments,and comparative experiments and ablation experiments are carried out by using the fault diagnosis accuracy rate as the model evaluation index.The results show that the average accuracy of MFCC-LSTM for low-speed gear fault diagnosis is as high as 99.14%,which is better than that of the comparison method,and MFCC-LSTM has a better recognition effect on low-speed gear faults.关键词
低速齿轮/故障诊断/梅尔倒谱系数/长短期记忆网络Key words
low-speed gears/fault diagnosis/Mel-frequency cepstrum coefficient/long short-term memory network分类
机械制造引用本文复制引用
张敬超,胡皓,李晨辉,宋金华,江国乾,李英伟..基于MFCC-LSTM的低速齿轮故障诊断方法研究[J].燕山大学学报,2025,49(5):404-413,10.基金项目
河北省军民融合科技创新计划资助项目(SJMYF202322) （SJMYF202322）
河北省重点实验室项目(202250701010046) （202250701010046）
