Small-sample and imbalanced milling chatter detection: Improved GAN with attention and hybrid deep learning

Haining  Gao; Xinli  Xiong; Hongdan  Shen; Yong  Yang; Yinlin  Wang

doi:10.59400/sv3502

Small-sample and imbalanced milling chatter detection: Improved GAN with attention and hybrid deep learning

Haining Gao Henan Province International Joint Laboratory of New Energy Digitalization Technology, Huanghuai University, Zhumadian 463000, China; School of Mechanical and Power Engineering, Hennan Polytechnic University, Jiaozuo 454000, China
Xinli Xiong Henan Province International Joint Laboratory of New Energy Digitalization Technology, Huanghuai University, Zhumadian 463000, China
Hongdan Shen Henan Province International Joint Laboratory of New Energy Digitalization Technology, Huanghuai University, Zhumadian 463000, China
Yong Yang Henan Province International Joint Laboratory of New Energy Digitalization Technology, Huanghuai University, Zhumadian 463000, China
Yinlin Wang Henan Province International Joint Laboratory of New Energy Digitalization Technology, Huanghuai University, Zhumadian 463000, China

Article ID: 3502

Keywords: milling chatter detection; small-sample learning; imbalanced data; improved GAN; attention mechanism; hybrid deep learning; data augmentation; machining monitoring

Abstract

Chatter detection during milling processes plays a pivotal role in ensuring machining quality and efficiency. While the accuracy of chatter detection heavily relies on experimental data, systems tend to exhibit overfitting phenomena under conditions of limited training samples, resulting in diminished detection precision. To address this limitation, this study presents a data augmentation algorithm based on an Improved Generative Adversarial Network (IGAN). This algorithm integrates advanced techniques including Wasserstein distance metrics, cycle consistency constraints, and channel attention mechanisms, effectively enhancing the quality of generated data. An innovative milling chatter detection deep learning model (MNBGA) is constructed, synthesizing cutting-edge architectures such as multi-scale convolutional neural networks, bidirectional gated recurrent neural networks, and attention mechanisms. To optimize model performance, the Ivy algorithm is employed for hyperparameter optimization of the MNBGA model. When the training dataset comprises 40 or more samples, the proposed method achieves detection accuracy exceeding 90%. Notably, under extreme imbalanced data conditions (24:1:1 ratio), the detection accuracy maintains 84.32%. The processing time for 40 samples requires only 76.17 ms, meeting real-time monitoring requirements. This research presents a novel technical solution for addressing the challenge of milling chatter detection under small-sample conditions.

Published

2025-06-15

How to Cite

Gao, H., Xiong, X., Shen, H., Yang, Y., & Wang, Y. (2025). Small-sample and imbalanced milling chatter detection: Improved GAN with attention and hybrid deep learning. Sound & Vibration, 59(3), 3502. https://doi.org/10.59400/sv3502

Download Citation

Issue

Vol. 59 No. 3 (2025)

Section

Article

This work is licensed under a Creative Commons Attribution 4.0 International License.

References

[1]Wang R, Song Q, Liu Z, et al. Multi-condition identification in milling Ti-6Al-4V thin-walled parts based on sensor fusion. Mechanical Systems and Signal Processing. 2022; 164: 108264. doi: 10.1016/j.ymssp.2021.108264

[2]Gao H, Shen H, Liu X, et al. Mechanics and dynamics research considering the tool radial runout effect in plunge milling. The International Journal of Advanced Manufacturing Technology. 2019; 106(5-6): 2391-2402. doi: 10.1007/s00170-019-04780-1

[3]Cao H, Lei Y, He Z. Chatter identification in end milling process using wavelet packets and Hilbert–Huang transform. International Journal of Machine Tools and Manufacture. 2013; 69: 11-19. doi: 10.1016/j.ijmachtools.2013.02.007

[4]Liu Y, Wang X, Lin J, et al. An adaptive grinding chatter detection method considering the chatter frequency shift characteristic. Mechanical Systems and Signal Processing. 2020; 142: 106672. doi: 10.1016/j.ymssp.2020.106672

[5]Wan S, Liu S, Li X, et al. Milling chatter detection based on information entropy of interval frequency. Measurement. 2023; 220: 113328. doi: 10.1016/j.measurement.2023.113328

[6]Chen D, Zhang X, Zhao H, et al. Development of a novel online chatter monitoring system for flexible milling process. Mechanical Systems and Signal Processing. 2021; 159: 107799. doi: 10.1016/j.ymssp.2021.107799

[7]Wang C, Zhang Y, Hu J. Adaptive milling chatter identification based on sparse dictionary considering noise estimation and critical bandwidth analysis. Journal of Manufacturing Processes. 2023; 106: 328-337. doi: 10.1016/j.jmapro.2023.10.012

[8]Wan M, Wang WK, Zhang WH, et al. Chatter detection for micro milling considering environment noises without the requirement of dominant frequency. Mechanical Systems and Signal Processing. 2023; 199: 110451. doi: 10.1016/j.ymssp.2023.110451

[9]Lu Y, Ma H, Sun Y, et al. An interpretable anti-noise convolutional neural network for online chatter detection in thin-walled parts milling. Mechanical Systems and Signal Processing. 2024; 206: 110885. doi: 10.1016/j.ymssp.2023.110885

[10]Xiao C, Wang L, Yu J. Order-difference of normalized square envelope spectrum and its applications in early chatter detection of roll grinder. Mechanical Systems and Signal Processing. 2025; 224: 112067. doi: 10.1016/j.ymssp.2024.112067

[11]Zhang X, Zheng L, Fan W, et al. Multi-domain data-driven chatter detection in robotic milling under varied robot poses based on directional attention mechanism. Mechanical Systems and Signal Processing. 2025; 227: 112406. doi: 10.1016/j.ymssp.2025.112406

[12]Deng K, Yang L, Lu Y, et al. Multitype chatter detection via multichannelinternal and external signals in robotic milling. Measurement. 2024; 229: 114417. doi: 10.1016/j.measurement.2024.114417

[13]Gao H, Wang H, Shen H, et al. Multi-modal denoised data-driven milling chatter detection using an optimized hybrid neural network architecture. Scientific Reports. 2025; 15(1). doi: 10.1038/s41598-025-88242-7

[14]Albertelli P, Braghieri L, Torta M, et al. Development of a generalized chatter detection methodology for variable speed machining. Mechanical Systems and Signal Processing. 2019; 123: 26-42. doi: 10.1016/j.ymssp.2019.01.002

[15]Zhao Y, Adjallah KH, Sava A, et al. Incipient chatter fast and reliable detection method in high-speed milling process based on cumulative strategy. ISA Transactions. 2022; 131: 397-414. doi: 10.1016/j.isatra.2022.05.039

[16]Matthew D E, Shi J, Hou M, et al. Improved STFT analysis using time-frequency masking for chatter detection in the milling process. Measurement, 2024, 225: 113899. doi: 10.1016/j.measurement.2023.113899.

[17]Yang B, Guo K, Zhou Q, et al. Early chatter detection in robotic milling under variable robot postures and cutting parameters. Mechanical Systems and Signal Processing. 2023; 186: 109860. doi: 10.1016/j.ymssp.2022.109860

[18]Shi J, Matthew DE, Tian W, et al. Adaptive threshold discrimination and synchronous squeezing transform for high-speed milling chatter detection. Journal of Manufacturing Processes. 2024; 131: 619-640. doi: 10.1016/j.jmapro.2024.09.030

[19]Wan S, Li X, Yin Y, et al. Milling chatter detection by multi-feature fusion and Adaboost-SVM. Mechanical Systems and Signal Processing. 2021; 156: 107671. doi: 10.1016/j.ymssp.2021.107671

[20]Zhang P, Gao D, Hong D, et al. Improving generalisation and accuracy of on-line milling chatter detection via a novel hybrid deep convolutional neural network. Mechanical Systems and Signal Processing. 2023; 193: 110241. doi: 10.1016/j.ymssp.2023.110241

[21]Sun Y, He J, Ma H, et al. Online chatter detection considering beat effect based on Inception and LSTM neural networks. Mechanical Systems and Signal Processing. 2023; 184: 109723. doi: 10.1016/j.ymssp.2022.109723

[22]Li R, Wei P, Liu X, et al. Cutting tool wear state recognition based on a channel-space attention mechanism. Journal of Manufacturing Systems. 2023; 69: 135-149. doi: 10.1016/j.jmsy.2023.06.010

[23]Wei P, Li R, Liu X, et al. Research on tool wear state identification method driven by multi-source information fusion and multi-dimension attention mechanism. Robotics and Computer-Integrated Manufacturing. 2024; 88: 102741. doi: 10.1016/j.rcim.2024.102741

[24]Gao H, Shen H, Yue C, et al. A monitoring method of milling chatter based on optimized hybrid neural network with attention mechanism. Facta Universitatis; 2024.

[25]Jauhari K, Rahman AZ, Huda MA, et al. A hybrid deep learning-based approach for on-line chatter detection in milling using deep stem-inception networks and residual channel-spatial attention mechanisms. Mechanical Systems and Signal Processing. 2025; 226: 112357. doi: 10.1016/j.ymssp.2025.112357

[26]Wang Y, Gao X, Wang P, et al. Multi-scale enhancement and chatter prediction of milling vibration signals from thin-walled parts based on attentional representation enhancement network. Mechanical Systems and Signal Processing. 2025; 225: 112302. doi: 10.1016/j.ymssp.2025.112302

[27]Sun H, Jin H, Zhuo Y, et al. Investigation on a chatter detection method based on meta learning for machining multiple types of workpieces. Journal of Manufacturing Processes. 2024; 131: 1815-1832. doi: 10.1016/j.jmapro.2024.09.091

[28]Liu C, Xu X, Wu J, et al. Deep transfer learning-based damage detection of composite structures by fusing monitoring data with physical mechanism. Engineering Applications of Artificial Intelligence. 2023; 123: 106245. doi: 10.1016/j.engappai.2023.106245

[29]Zhang X, Shi B, Feng B, et al. A hybrid method for cutting tool RUL prediction based on CNN and multistage Wiener process using small sample data. Measurement. 2023; 213: 112739. doi: 10.1016/j.measurement.2023.112739

[30]Zhao S, Zhang H, Peng F, et al. United optimization strategy of ultrasonic vibration assisted process and multiple parameters for machining deformation reduction. Journal of Manufacturing Processes. 2024; 131: 1942-1958. doi: 10.1016/j.jmapro.2024.09.111

[31]Zhao S, Peng F, Sun H, et al. Physical multi-factor driven nonlinear superposition for machining deformation reconstruction. International Journal of Mechanical Sciences. 2024; 262: 108723. doi: 10.1016/j.ijmecsci.2023.108723

[32]Yin C, Wang Y, Ko JH, et al. Attention-driven transfer learning framework for dynamic model guided time domain chatter detection. Journal of Intelligent Manufacturing. 2023; 35(4): 1867-1885. doi: 10.1007/s10845-023-02133-0

[33]Jiang P, Fan J, Li L, et al. A hybrid approach combining mechanism-guided data augmentation and machine learning for biomass pyrolysis. Chemical Engineering Science. 2024; 296: 120227. doi: 10.1016/j.ces.2024.120227

[34]Kuo PH, Luan PC, Tseng YR, et al. Machine tool chattering monitoring by Chen-Lee chaotic system-based deep convolutional generative adversarial nets. Structural Health Monitoring. 2023; 22(6): 3891-3907. doi: 10.1177/14759217231159865

[35]Arjovsky M, Chintala S, Bottou L. Wasserstein generative adversarial networks. International Conference on Machine Learning. 2017.

[36]Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of Wasserstein Gans. In: Proceedings of the 31st International Conference on Neural Information Processing Systems; 2017.

[37]Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets. In: Proceedings of the 28th International Conference on Neural Information Processing Systems; 2014.

[38]Lecun Y, Bottou L, Bengio Y, et al. Gradient-based learning applied to document recognition. Proceedings of the IEEE. 1998; 86(11): 2278-2324. doi: 10.1109/5.726791

[39]Ma J, Luo D, Liao X, et al. Tool wear mechanism and prediction in milling TC18 titanium alloy using deep learning. Measurement. 2021; 173: 108554. doi: 10.1016/j.measurement.2020.108554

[40]Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need. Advances in neural information processing systems. arXiv. 2017.

[41]Ghasemi M, Zare M, Trojovský P, et al. Optimization based on the smart behavior of plants with its engineering applications: Ivy algorithm. Knowledge-Based Systems. 2024; 295: 111850. doi: 10.1016/j.knosys.2024.111850

[42]Powers DMW. Evaluation: From precision, recall and F-measure to ROC, informedness, markedness and correlation. arXiv. 2020.

[43]Gao H, Shen H, Yu L, et al. Milling chatter detection system based on multi-sensor signal fusion. IEEE Sensors Journal. 2021; 21(22): 25243-25251. doi: 10.1109/jsen.2021.3058258

[44]Zhou J, Yue C, Qu J, et al. BDTM-Net: A tool wear monitoring framework based on semantic segmentation module. Journal of Manufacturing Systems. 2024; 77: 576-590. doi: 10.1016/j.jmsy.2024.10.012

This paper introduces a universal framework for understanding the vibration responses of systems subjected to harmonic excitation. By examining a simplified cylinder-spring-damper model, the study refurbishes traditional scaling methods for the excitation frequency ratio and critical damping ratio. The findings indicate that in damped systems, the maximum amplitude of vibration does not align with the natural frequency. This observation leads to the introduction of a new scaling method for reduced frequency. This new approach aligns resonance peaks at the new reduced velocity of 1.0 across different damping ratios, providing a consistent characterization of vibration behavior. A new critical damping ratio of 0.707 is identified for an excited system as opposed to the traditional damping ratio of 1.0 for an unexcited system. Key properties such as maximum amplitude, phase lag, bandwidth, and quality factor are analyzed, demonstrating that the proposed reduced frequency and critical damping ratio effectively capture the dynamics of both damped and undamped excited systems. The findings offer significant insights for practical applications in engineering and various scientific fields.

This study makes significant contributions to the field of ultrasonic testing (UT) by offering a novel approach to the identification of artificially introduced defects within Japanese cedar wood (Cryptomeria japonica). The findings are of particular relevance for the heritage conservation and construction sectors, where non-invasive defect detection is paramount. The study establishes a robust framework for assessing the structural integrity of timber by correlating ultrasonic wave velocity reductions with defect size and distribution. Big-sized defects led to more substantial decreases in wave velocity. The study establishes a robust framework for assessing the structural integrity of historical timber by correlating ultrasonic wave velocity reductions with defect size and distribution. This framework has the potential to be applicable to diverse wood species and defect types.

The control of vehicle interior noise has become a critical metric for assessing noise, vibration, and harshness (NVH) in vehicles. During the initial phases of vehicle development, accurately predicting the impact of road noise on interior noise is essential for reducing noise levels and expediting the product development cycle. In recent years, data-driven methods based on machine learning have gained significant attention due to their robust capability in navigating complex data mapping relationships. Notably, surrogate models have demonstrated exceptional performance in this domain. Numerous researchers have integrated diverse intelligent algorithms into the study of vehicle noise, leveraging advantages such as the elimination of precise modeling requirements, extensive solution space exploration, continuous learning from data, and robust algorithmic versatility. However, in NVH engineering applications, data-driven models face inherent limitations, particularly in interpretability and stability. To address these issues, this paper introduces an improved Long Short-Term Memory (LSTM) network that combines knowledge and data. Inspired by the physical information neural network concept, this approach incorporates values calculated through empirical formulas into the neural network as constraints. Comparative assessments with traditional LSTM networks highlight the advantages of this deep learning model. By integrating empirical formulas constraints, the model not only enhances interpretability but also achieves robust generalization with fewer data samples. The proposed method is validated on a specific vehicle model, showing significant improvements in prediction accuracy and efficiency.