梅尔频谱图驱动的深度学习框架：用于丝材电弧增材制造中基于声发射的制造监测

Fei Gao; Yishu Chen; Jing Lin; Yinmin Zhu; Yonghao Miao; Jinghui Tian

doi:10.1088/3050-2454/ae5498

您当前的位置：

首页 >

文章列表页 >

梅尔频谱图驱动的深度学习框架：用于丝材电弧增材制造中基于声发射的制造监测

Papers | 更新时间：2026-04-14

- 梅尔频谱图驱动的深度学习框架：用于丝材电弧增材制造中基于声发射的制造监测
- Mel spectrogram–driven deep learning framework for acoustic emission-based manufacturing monitoring in wire arc additive manufacturing
- 可靠性科学与工程学报（英文） 2026年页码：1-14
- 作者机构：
  
  School of Reliability and Systems Engineering, Beihang University, Xueyuan Road No.37, Haidian District, Beijing, People's Republic of China
  Ningbo Institution of Technology (NIT), Beihang University, Ningbo 315832, People's Republic of China
- 作者简介：
  
  [ "Fei Gao received BS and PhD degrees in mechanical engineering from Xi'an Jiaotong University, Xi'an, China, in 2013 and 2018, respectively. He is currently an Associate Professor with the School of Reliability and Systems Engineering, Beihang University, Beijing, China. His research interests include structural health monitoring (SHM), nondestructive testing (NDT), and guided wave propagation." ]
  [ "Yishu Chen received a BS degree in reliability and safety engineering from Beihang University, Beijing, China, in 2022. He is currently pursuing a PhD degree in control science and engineering at Beihang University, Beijing, China. His research interests include structural health monitoring (SHM), nondestructive testing (NDT), and digital twin for fault diagnosis." ]
  [ "Jing Lin (Senior Member, IEEE) received BS, MS and PhD degrees in mechanical engineering from Xi'an Jiaotong University, Xi'an, China, in 1993, 1996 and 1999, respectively. From 2009 to 2018, he was a Professor with the School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an, China. He is currently the Dean of the School of Reliability and Systems Engineering, Beihang University, Beijing, China." ]
  [ "Yinmin Zhu received a BS degree in safety engineering from Beihang University, Beijing, China, in 2024. He is currently pursuing a PhD degree in electronic information engineering with Beihang University, Beijing, China. His research interests include structural health monitoring (SHM), nondestructive testing (NDT), and intelligent signal processing technology." ]
  [ "Yonghao Miao received BS degree in mechanical engineering and automation from the Wuhan University of Technology, Wuhan, China, in 2013, and a PhD degree in mechanical engineering from Xi'an Jiaotong University, Xi'an, China, in 2018. He is currently an Associate Professor with the School of Reliability and Systems Engineering, Beihang University, Beijing, China." ]
  [ "Jinghui Tian received a BS degree in measurement and control technology and instrumentation from Henan Polytechnic University, Jiaozuo, China, in 2018, and a PhD degree in mechanical design and theory from Yanshan University, Qinhuangdao, China, in 2024. He was a Visiting PhD Student with the Department of Mechanical Engineering, Politecnico di Milano, Milan, Italy, from 2022 to 2023. He is currently a Postdoctoral Research Fellow with the Ningbo Institute of Technology, Beihang University, Ningbo, China. His research interests include transfer learning, information fusion, machinery condition monitoring, and intelligent fault diagnosis." ]
- 基金信息：
- DOI：10.1088/3050-2454/ae5498
  中图分类号：
- 收稿：2025-12-26，
  
  修回：2026-03-14，
  
  录用：2026-03-18，
  
  网络首发：2026-03-31，
  
  纸质出版：2026-06
- 稿件说明：
移动端阅览
Fei Gao, Yishu Chen, Jing Lin, et al. 梅尔频谱图驱动的深度学习框架：用于丝材电弧增材制造中基于声发射的制造监测[J]. 可靠性科学与工程学报（英文）, 2026, 2: 025301.

Fei Gao, Yishu Chen, Jing Lin, et al. Mel spectrogram–driven deep learning framework for acoustic emission-based manufacturing monitoring in wire arc additive manufacturing[J]. Journal of Reliability Science and Engineering, 2026, 2: 025301.
Fei Gao, Yishu Chen, Jing Lin, et al. 梅尔频谱图驱动的深度学习框架：用于丝材电弧增材制造中基于声发射的制造监测[J]. 可靠性科学与工程学报（英文）, 2026, 2: 025301. DOI： 10.1088/3050-2454/ae5498.

Fei Gao, Yishu Chen, Jing Lin, et al. Mel spectrogram–driven deep learning framework for acoustic emission-based manufacturing monitoring in wire arc additive manufacturing[J]. Journal of Reliability Science and Engineering, 2026, 2: 025301. DOI： 10.1088/3050-2454/ae5498.

摘要

声发射技术在丝材电弧增材制造（WAAM）的原位监测与缺陷诊断中展现出巨大潜力。然而，WAAM过程中产生的AE信号受到电弧放电噪声的严重污染，导致难以直接从原始数据中提取与缺陷相关的特征。本研究提出一种基于梅尔频谱图的深度学习框架，用于WAAM的高效健康监测。该方法引入了一种增强的时频表示策略，能够在WAAM工艺固有的高噪声条件下更有效地捕捉缺陷相关特征。通过利用梅尔频谱图表示，该框架增强了对信息丰富的低频部分的表示，同时抑制了高频噪声，从而提升了特征的可解释性和鲁棒性。研究采用卷积神经网络和视觉变换器模型进行缺陷分类及性能基准测试。实验结果表明，所提方法在显著降低计算成本的同时实现了较高的诊断精度，优于传统的基于短时傅里叶变换的方法。研究结果证实，梅尔频谱图表示为WAAM的健康监测提供了一种更高效且泛化能力更强的解决方案。

Abstract

Acoustic emission shows great potential for

in-situ

monitoring and defect diagnosis in wire arc additive manufacturing (WAAM). However

the AE signal produced during WAAM is heavily contaminated by arc discharge noise

making it difficult to directly extract defect-related features from raw data. This study proposes a Mel spectrogram–based deep learning framework for efficient WAAM health monitoring. The method introduces an enhanced time–frequency representation strategy that captures defect-related features more effectively under the high-noise conditions inherent in WAAM processes. By leveraging Mel spectrogram representations

the framework emphasizes informative low-frequency components while suppressing high-frequency noise

thereby improving feature interpretability and robustness. Convolutional neural network and vision transformer models are employed for defect classification and performance benchmarking. Experimental results demonstrate that the proposed approach achieves high diagnostic

accuracy with substantially reduced computational cost

outperforming conventional short-time Fourier transform-based methods. The findings confirm that Mel spectrogram representations offer a more efficient and generalizable solution for health monitoring in WAAM.

关键词

Keywords

references

Gain S and Veeman D 2025 A review on advances and challenges in wire arc additive manufacturing: process parameters, microstructural evolution and material performance across alloys J. Alloys Compd. 1029 180735

Shah A , Aliyev R , Zeidler H and Krinke S 2023 A review of the recent developments and challenges in wire arc additive manufacturing (WAAM) process J. Manuf. Mater. Process. 7 97

Ding C , Wang J , Zhang S , Yang S and Ma Y 2025 A novel active learning stochastic Kriging metamodel for improving reliability and stability of additive manufacturing processes Reliab. Eng. Syst. Saf. 260 111043

Wu B , Pan Z , Ding D , Cuiuri D , Li H , Xu J and Norrish J 2018 A review of the wire arc additive manufacturing of metals: properties, defects and quality improvement J. Manuf. Process. 35 127 – 39

Thompson A , Maskery I and Leach R 2016 X-ray computed tomography for additive manufacturing: a review Meas. Sci. Technol. 27 072001

Bhat G A , Smagulova D and Jasiunienė E 2026 Optimization of nondestructive evaluation techniques for bonded components through model-assisted POD analysis IEEE Trans. Reliab. 75 555 – 69

Choudhary A , Goyal D and Letha S S 2021 Infrared thermography-based fault diagnosis of induction motor bearings using machine learning IEEE Sens. J. 21 1727 – 34

Mulaveesala R , Arora V and Dua G 2021 Pulse compression favorable thermal wave imaging techniques for non-destructive testing and evaluation of materials IEEE Sens. J. 21 12789 – 97

Vilar N , Artigas R , Bermudez C , Thompson A , Newton L , Leach R , Duocastella M and Carles G 2022 Optical system for the measurement of the surface topography of additively manufactured parts Meas. Sci. Technol. 33 104001

Li C , Guo H , Wei X , He W , Nie X , Zhang T and Fang Y 2025 SaimVAE: an unlabeled intelligent ultrasonic NDT method for composite materials Measurement 249 117059

Wei W , Liu Y , Wu J , Wei Z , Zhou Z and Long Y 2025 In-situ monitoring method of femtosecond laser welding between glass and copper with acoustic emission Measurement 240 115568

He J , Yang F , Wang H , Sun X , Zhu Y , Wang Y and Guan X 2025 A physics-based acoustic emission energy method for mixed-mode impact damage prediction of composite laminates Ultrasonics 145 107490

Wang X , Yue Q and Liu X 2024 Reliable arrival time picking of acoustic emission using ensemble machine learning models Mech. Syst. Signal Process. 215 111442

Zhao S , Li G and Wang C 2024 Bridge cable damage identification based on acoustic emission technology: a comprehensive review Measurement 237 115195

Zhang Z , Huang W , Liao Y , Song Z , Shi J , Jiang X , Shen C and Zhu Z 2022 Bearing fault diagnosis via generalized logarithm sparse regularization Mech. Syst. Signal Process. 167 108576

Qiu T , Huang W , Zhang Z , Wang J and Zhu Z 2024 A new approach for sparse optimization with Moreau envelope to extract bearing fault feature Mech. Syst. Signal Process. 216 111493

Jiao J , Zhang J , Ren Y , Li G , Wu B and He C 2023 Sparse representation of acoustic emission signals and its application in pipeline leak location Measurement 216 112899

Subasi L , Oren S , Dursun G , Sen C and Orhangul A 2024 In-situ acoustic signals correlation of process parameters in laser powder bed fusion Proc. CIRP. 126 668 – 73

Gao Z , Lin J , Wang X and Xu X 2017 Bearing fault detection based on empirical wavelet transform and correlated kurtosis by acoustic emission Materials 10 571

Cui J , Qu X , Lv C , Du J and Wang H 2025 Multi-parameter acoustic emission analysis for fatigue crack evaluation in structural health monitoring Measurement 256 118529

Gao F , Li C , Ji D , Hua J and Lin J 2025 Spectral characterization method for wire-arc additive manufacturing monitoring with broadband AE signals Meas. Sci. Technol. 36 035002

Zhuang J , Wu J , Pei G , Li W , Xin G , Ma C , Feng K , Zhang L and Yang C 2025 A neuro-fuzzy approach with hypergraph convolution for fault diagnosis in industrial devices J. Reliab. Sci. Eng. 1 035301

Ju S , Li D and Jia J 2022 Machine-learning-based methods for crack classification using acoustic emission technique Mech. Syst. Signal Process. 178 109253

Liu K , Wulan T , Yao Y , Bian M and Bao Y 2024 Assessment of damage evolution of concrete beams strengthened with BFRP sheets with acoustic emission and unsupervised machine learning Eng. Struct. 300 117228

Ji D , Li W , Liu Z , Liu H , Gao F , Wang M , Chang B and Lin J 2025 Quantitative stiffness evaluation and damage mechanism investigation in open-hole composites via multi-modal SHM data fusion Mech. Syst. Signal Process. 237 112998

Abdul Z K and Al-Talabani A K 2022 Mel frequency cepstral coefficient and its applications: a review IEEE Access 10 122136 – 58

Ren X , Wang J , Liang Y , Ma L and Zhou W 2024 Acoustic emission detection of filament wound CFRP composite structure damage based on Mel spectrogram and deep learning Thin-Walled Struct. 198 111683

Yang F , Li Z , Liu S , He X , Song D , Li N , Wang H and Sobolev A 2025 Characterizing the fracture evolution of sandstone by using Mel-frequency cepstral coefficients of acoustic emission Eng. Fract. Mech. 324 111254

Alzubaidi L , Zhang J , Humaidi A J , Al-Dujaili A , Duan Y , Al-Shamma O , Santamaría J , Fadhel M A , Al-Amidie M and Farhan L 2021 Review of deep learning: concepts, CNN architectures, challenges, applications, future directions J. Big Data 8 53

Hamed Alizadeh Moghaddam S , Gazor S , Homayouni S and Karami F 2025 Integrating local and global features in a convolutional vision transformer for wetland mapping IEEE Sens. J. 25 8674 – 83

Dosovitskiy A et al 2021 An image is worth 16x16 words: transformers for image recognition at scale (arXiv:2010.11929)

Qu L , Li X , Yang T and Wang S 2025 Radar-based continuous human activity recognition using multidomain fusion vision transformer IEEE Sens. J. 25 9946 – 56

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

暂无数据