CNN-powered malfunction detection in industrial machine sounds with fair data splitting and dynamic training control

Ameer B.A. Alaasam; Zena H. Khalil; Dhiah Al-Shammary; Ayman Ibaida; Khandakar Ahmed

doi:10.1016/j.jnlest.2026.100372

Journal of Electronic Science and Technology ›› 2026, Vol. 24 ›› Issue (3) :100372 DOI: 10.1016/j.jnlest.2026.100372

Regular Papers

research-article

CNN-powered malfunction detection in industrial machine sounds with fair data splitting and dynamic training control

Author information +

¹ School of Electronic Engineering and Computer Science, South Ural State University, Chelyabinsk, 454080, Russia

² College of Computer Science and Information Technology, University of Al-Qadisiyah, Al Diwaniyah, 58001, Iraq

³ Directorate for Education in Al Qadisiyah under the Ministry of Education of Iraq, Al Diwaniyah, 58001, Iraq

⁴ Intelligent Technology Innovation Lab, Victoria University, Melbourne, 3011, Australia

^* E-mail addresses: alaasamab@susu.ru (A.B. Alaasam),

zena.khalil@qu.edu.iq (Z.H. Khalil),

d.alshammary@qu.edu.iq (D. Al-Shammary),

ayman.ibaida@vu.edu.au (A. Ibaida),

khandakar.ahmed@vu.edu.au (K. Ahmed).

Ameer B.A. Alaasam received the Ph.D. degree in computer science from South Ural State University, Chelyabinsk, Russia in 2022. He is currently an Associate Professor with the Department of System Programming, School of Electronic Engineering and Computer Science, South Ural State University. He has been involved in teaching and scientific research in computer science since 2007. He has participated in multiple funded research projects in Russia focusing on fog computing, digital twins, and containerized cloud solutions. He developed registered platforms, including CLOUDFOGNET and a micro-workflow concept implemented in the Kepler Scientific Workflow platform. His research interests include stream processing, distributed and parallel computing, digital twins, and smart industry technologies

Zena H. Khalil received the B.Sc. degree in computer science from the University of Al-Qadisiyah, Al Diwaniyah, Iraq, and the M.Sc. degree from Babylon University, Hillah, Iraq in 2005 and 2011, respectively. In 2021, she completed postgraduate studies from South Ural State University. Since 2005, she has been employed as a Lecturer with the University of Al-Qadisiyah. She also worked as a Research Engineer with South Ural State University from late 2018 to late 2020. Her research interests include signal and digital image processing, remote sensing, artificial intelligence (AI), machine learning, and biometrics

Dhiah Al-Shammary obtained the Ph.D. degree in computer science from RMIT University, Melbourne, Australia in 2014. Currently he is a Senior Academic at the University of Al-Qadisiyah. He worked for several universities in both Iraq and Australia. He worked in some Silicon Valley based companies in Australia like Optcts and AgilePQ. He has published with his Australian-USA team a potential patent for designing new post-quantum data encryption with high performance. He has several publications at highly reputed venues. His special interests include data security and privacy, clustering techniques, classifications methods, optimization, and networking

Ayman Ibaida received the Ph.D. degree in computer science and IT from RMIT University in 2014. He was a Lecturer with the Computer College, United Arab Emirates University, Al Ain, Dubai from 2006 to 2008. He was the Technical Lead at AgilePQ Australia Ltd. and a Cofounder of EyeCura Pty Ltd., Melbouen, Australia in 2017. He is currently a Lecturer with Victoria University, Melbourne, Australia. His research interests include AI and machine learning in biomedical applications, diagnoses, and patient health records security as well as cyber security in healthcare systems

Khandakar Ahmed received the Ph.D. degree from RMIT University in 2015 and the M.Sc. degree in networking and e-business centered computing (NeBCC) under the joint consortia of the University of Reading, Reading, UK in 2009; Aristotle University of Thessaloniki, Thessaloniki, Greece, and Charles III University of Madrid (UC3M), Madrid, Spain in 2011. He started the career as a Full-Stack Developer and later switched to the academy by joining his alma mater as a Lecturer in 2007, when he completed his undergraduate program. He taught at several universities in Europe and Australia and worked as a Post-Doctoral Research Fellow at Royal Melbourne Institute of Technology (RMIT) from 2015 to 2016, before joining Victoria University in 2017. He is currently a Senior Lecturer with the College of Engineering and Science, Victoria University. He has extensive industry engagement as a Chief Investigator in multiple research projects related to the Internet of things, smart cities, machine learning, cybersecurity, and biomedical informatics

Show less

History +

PDF (7527KB)

Abstract

In an industrial environment, sound is generally an important source of data that results from the movement and vibration of machines during operation. Each machine operation condition often generates a specific sound, for example, a familiar normal working sound, an unfamiliar sound, a breaking sound, or even the absence of sound, which may indicate a stop. This paper proposes a supervised convolutional neural network (CNN) model with early stopping and learning rate (LR) scheduling to achieve a more accurate and cost-effective trained deep learning (DL) model for machine sound classification. The proposed model is trained and evaluated using the MIMII dataset, which contains unbalanced real-world data and signal-to-noise ratios (SNRs). To ensure a fair (i.e., representative) data splitting strategy and robust coverage across all hierarchical structures of the MIMII dataset, this study adopts a hierarchical multi-stage stratified splitting protocol incorporating pre-split shuffling to prevent order bias. The synthetic minority oversampling technique (SMOTE) is used to balance the training set after feature extraction. Compared with state-of-the-art works implemented on the same dataset, the proposed model achieves optimal performance in terms of accuracy, area under the curve (AUC) , F1-score, precision, recall, Matthews correlation coefficient (MCC), and Kappa without noise processing methods at an SNR of 6 dB, and maintains robust, competitive performance under severe noise conditions (0 dB and −6 dB).

Keywords

Convolutional neural network / Early stopping / Imbalanced data classification / Industrial machine sound / Learning rate scheduling / Malfunction detection

Cite this article

Download citation ▾

Ameer B.A. Alaasam, Zena H. Khalil, Dhiah Al-Shammary, Ayman Ibaida, Khandakar Ahmed. CNN-powered malfunction detection in industrial machine sounds with fair data splitting and dynamic training control. Journal of Electronic Science and Technology, 2026, 24 (3) : 100372 DOI:10.1016/j.jnlest.2026.100372

登录浏览全文

4963

注册一个新账户忘记密码

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	R.V. Sharan, H. Xiong, S. Berkovsky, Benchmarking audio signal representation techniques for classification with convolutional neural networks, Sensors 21 (10) (2021) 3434.

[2]	L. Hertel, H. Phan, A. Mertins, Comparing time and frequency domain for audio event recognition using deep learning, in: Proc. of the 2016 Intl. Joint Conf. on Neural Networks (IJCNN), Vancouver, Canada, 2016, pp. 3407-3411.

[3]	P. Golik, Z. Tüske, R. Schlüter, et al., Convolutional neural networks for acoustic modeling of raw time signal in LVCSR, in: Proc. of the INTERSPEECH 2015, 16th Annual Conf. of the Intl. Speech Communication Association, Dresden, Germany, 2015, pp. 26-30.

[4]	T. Sercu, C. Puhrsch, B. Kingsbury, et al., Very deep multilingual convolutional neural networks for LVCSR, in: Proc. of the 2016 IEEE Intl. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Shanghai, China, 2016, pp. 4955-4959.

[5]	A. Graves, A.R. Mohamed, G. Hinton, Speech recognition with deep recurrent neural networks, in: Proc. of the 2013 IEEE Intl. Conf. on Acoustics, Speech and Signal Processing, Vancouver, Canada, 2013, pp. 6645-6649.

[6]	D.S. Park, W. Chan, Y. Zhang, et al., SpecAugment: a simple data augmentation method for automatic speech recognition, in: Proc. of the 20th Annual Conf. of the Intl. Speech Communication Association, Graz, Austria, 2019, pp. 2613-2617.

[7]	V. Pann, K.S. Kwon, B. Kim, et al., Robustness of CNN-based model assessment for pig vocalization classification across diverse acoustic environments, Comput. Electron. Agric. 240 (2026) 111181.

[8]	A.B.A. Alaasam, G.I. Radchenko, A.N. Tchernykh, Micro-workflows data stream processing model for industrial internet of things, Supercomput, Front. Innov. 8 (1) (2021) 82-98.

[9]	J. Konar, P. Khandelwal, R. Tripathi, Comparison of various learning rate scheduling techniques on convolutional neural network, in: Proc. of the 2020 IEEE Intl. Students’ Conf. on Electrical, Electronics and Computer Science (SCEECS), Bhopal, India, 2020, pp. 1-5.

[10]	A. Fernández, S. García, F. Herrera, et al., SMOTE for learning from imbalanced data: progress and challenges, marking the 15-year anniversary, J. Artif. Intell. Res., (1) (2018) 863-905.

[11]	Y.-H. Chen, Z.-Q. Xie, J.-Y. Zhong, et al., SQKformer: spiking sparse QKformer with adaptive batch normalization for membrane potential, Neurocomputing 671 (2026) 132666.

[12]	X.-Y. Jin, Analysis of residual block in the ResNet for image classification, in: Proc. of the 1st Intl. Conf. on Data Analysis and Machine Learning, Kuala Lumpur, Malaysia, 2023, pp. 246-250.

[13]	W. Han, C.F. Chan, C.S. Choy, et al., An efficient MFCC extraction method in speech recognition, in: Proc. of the 2006 IEEE Intl. Symposium on Circuits and Systems, Kos, Greece, 2006, p. 4.

[14]	M. Maseri, M. Mamat, Malay language speech recognition for preschool children using hidden Markov model (HMM) system training, in: Proc. of the 5th ICCST 2018 on Computational Science and Technology, Kota Kinabalu, Malaysia, 2019, pp. 205-214.

[15]	N. Madhu, Note on measures for spectral flatness, Electron. Lett. 45 (23) (2009) 1195-1196.

[16]	J. Herre, E. Allamanche, O. Hellmuth, Robust matching of audio signals using spectral flatness features, in: Proc. of the 2001 IEEE Workshop on the Applications of Signal Processing to Audio and Acoustics (Cat. No.01TH8575), New Platz, USA, 2001, pp. 127-130.

[17]	T. Giannakopoulos, A. Pikrakis, Audio features, T. Giannakopoulos, A. Pikrakis (Eds.), Introduction to Audio Analysis, Elsevier, Amsterdam, The Netherlands, 2014, pp. 59-103.

[18]	L. Gantert, M. Sammarco, M. Detyniecki, et al., A supervised approach for corrective maintenance using spectral features from industrial sounds, in: Proc. of the 2021 IEEE 7th World Forum on Internet of Things (WF-IoT), New Orleans, USA, 2021, pp. 723-728.

[19]	L. Gantert, M. Sammarco, M. Detyniecki, et al., Super learner ensemble for sound classification using spectral features, in: Proc. of the 2022 IEEE Latin-American Conf. on Communications (LATINCOM), Rio de Janeiro, Brazil, 2022, pp. 1-6.

[20]	B.A. Tama, M. Vania, I. Kim, et al., An EfficientNet-based weighted ensemble model for industrial machine malfunction detection using acoustic signals, IEEE Access 10 (2022) 34625-34636.

[21]

F. Qiu, S. Chen, D. Jin, et al., Fan sound fault detection algorithm based on the fusion of convolutional neural network and gated recurrent unit network, in: Proc. of the Sixth Intl. Conf. on Advanced Electronic Materials, Computers, and Software Engineering (AEMCSE 2023), Shenyang, China, 2023, p. 127870P.

[22]	L. Gantert, T. Zeffiro, M. Sammarco, et al., Multiclass classification of faulty industrial machinery using sound samples, Eng. Appl. Artif. Intell. 136 (2024) 108943.