Abstract
Percussion detection is a rapid screening method for materials defects commonly used in the inspection of boilers, pressure vessels and other special equipments, however, it is prone to misjudgment because of its dependence on personal experience. In this paper, percussion detection experiments were designed for 10 categories of samples and frequency features as well as the MFCC features were studied. A 13-layer two-channel convolutional neural network (CNN) was trained by a training sample generalized with Gauss random noises. The research results showed that the structural integrity of metal component was closely related to its percussion information, for instance, defects of flat components would reduce their percussion vibration frequency while cracks of straight pipe components would cause splitting of peaks in frequency domain. MFCC parameters of 40-dimensional percussion sound and percussion vibration were used as the input of the13-layer model, and the training samples were generalized with a certain intensity of random noise. The proposed CNN model coordinated the range and detail of feature extraction suitably, whose recognition rates of 10 categories of defects were more than 99% under condition of 5dB signal-to-noise ratio, and was also robust to periodic noise interference from air compressors. The test results showed the model was better than other CNN models with different network structures and was more robust compared with the traditional hidden Markov model. A 13-layer single channel CNN model was more suitable for practical application although the recognition rate of single channel model was lower than that of the two-channel model, and the effect of percussion sound was better than that of percussion vibration. Compared with the traditional vibration mode or time-frequency domain analysis methods, the method in this paper relied less on signal processing skills and expert knowledge and experience, and therefore could be easily used for percussion detection intellectualization.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Qiqi, Z., Banfu, Y.: Impact acoustic-based identification method for detecting interfacial debonding of steel-concrete composite slab. J. Railway Sci. Eng. 15(003), 668–676 (2018)
Jianping, J., Xiangpeng, L., Haijun, Z., et al.: Coating quality detection method on nonmetal substrate surface. Appl. Acoust. 36(2), 105–110 (2017)
Shaozheng, Z., Ning, N., Pengfei, Y., et al.: Composite sandwich structure digital tap test technology research. Aeronaut. Manuf. Technol. 61(3), 90–94 (2018)
Fei, Z.: Application of audio frequency method in vermicular graphite cast iron inspection. Foundry Technol. 6, 573–576 (2019)
Rongmei, L., Liang, S., Keying, Z.: Influence of impact hammer on parameter acoustic vibration detecting depth inside wind turbine blade. Nondestruct. Test. 38(5), 5–9 (2016)
Bowu, S., Lizhen, M., Jiaqi, Z.: An inspection hammer capable of percussion in narrow space for boiler inspection type. China Patent, 201921754643.X.2020-06-09
General Administration of Quality Supervision, Inspection and Quarantine, Standardization Administration of China. GB/T13004-2016 Periodic inspection and evaluation of seamless steel gas cylinders. Quality inspection of China press, Beijing (2016)
Yangchao, Z., Yanjun, L., Yuyuan, C., et al.: Study on percussion detection of aerospace composite materials based on k-means clustering. Aeronaut. Sci. Technol. (10), 45–49 (2019)
Yong, Y., Junwei., Rongzhe, L., et al.: Analysis on edge estimation of functional layer defect in ballastless track based on burg power spectrum. J. Vibrat. Shock 39(10), 1–6 (2020)
Sun, D., Yan, Y., Carter, R.M., et al.: On-line nonintrusive detection of wood pellets in pneumatic conveying pipelines using vibration and acoustic sensors. IEEE Trans. Instrum. Meas. 63(5), 993–1001 (2014)
Kotb, A., Abdulaziz, A.M.: Cavitation detection in variable speed pump by analyzing the acoustic and vibration spectrums. Engineering 07(10), 706–716 (2015)
Na, Z., Caixue, L., Taibo, Y., et al.: Application of VMD and random forest in reactor metal impact signal recognition. J. Vibrat. Shock 40(5), 102–107 (2021)
Junyu, Z., Yanming, Z.: Application of convolution neural network in image classification and object detection. Comput. Eng. Appl. 53(013), 34–41 (2017)
Jianhu, Y., Tao, H., Jian, T., et al.: An approach to intelligent fault diagnosis of rolling bearing using wavelet time-frequency representations and CNN. Mach. Des. Res. 02, 101–105 (2017)
Chunzhi, W., Pengcheng, J., Fuzhou, F., et al.: Faults diagnosis method for gearboxes based on a 1-D convolutional neural network. J. Vibrat. Shock 37(22), 51–56 (2018)
Ning, F., Yu, Z., Qingwei, Y., Yugang, L.: Modal parameter identification of vibration signal based on unsupervised learning convolutional neural network. J. Comput. Applicat. (03), 786–790 (2017)
Guoqiang, Z., Xiao, Z., Wentao, T.: Internal defects detection of brak pads based on the percussioning signal. Modular Mach. Tool Automat. Manuf. Tech. 10, 122–124 (2017)
Le, L., Wenfeng, L.: Study on model identification method using sound. Sci. Technol. Eng. 10(022), 5441–5444 (2010)
Xu, Z.T.: Research on Digital Coin-Tap Test’s Application Technology in Aerospace Composite Detection. Nanjing University of Aeronautics and Astronautics, Nanjing (2016)
Cao, H., Xu, C., Zhao, X., et al.: The Mel-frequency cepstral coefficients in speaker recognition. J. Northwest Univ. (Nat. Sci. Edition) 43(2), 203–208 (2013)
Wenhao, Y., Yingxi, L., Chunyan, L., et al.: Improving generalization ability of speech enhancement approaches using generated noise. Acta Electron. Sin. 47(4), 791–797 (2019)
Xinmin, L., Jing, Q., Guanjun, L.: Continuous Gaussian mixture HMM- based diagnosing method of roller bearing. J. Mech. Transmiss. 29(001), 7–10 (2005)
Lim, C.L.P., Woo, W.L., Dlay, S.S., et al.: Heartrate-dependent heartwave biometric identification with thresholding-based GMM–HMM methodology. IEEE Trans. Industr. Inf. 15(1), 45–53 (2019)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 Science Press
About this paper
Cite this paper
Yu, H., Zhao, X., Tang, Y., Chen, X., Du, X. (2022). Researches on Percussion Detection of Metal Component Based on Convolutional Neural Networks. In: Duan, B., Umeda, K., Kim, Cw. (eds) Proceedings of the Eighth Asia International Symposium on Mechatronics. Lecture Notes in Electrical Engineering, vol 885. Springer, Singapore. https://doi.org/10.1007/978-981-19-1309-9_56
Download citation
DOI: https://doi.org/10.1007/978-981-19-1309-9_56
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-19-1308-2
Online ISBN: 978-981-19-1309-9
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)