|Table of Contents|

Damage state assessment method for pinch roll based on multi-domain information deep fusion(PDF)

长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

Issue:
2026年01期
Page:
189-198
Research Field:
汽车与机械工程
Publishing date:

Info

Title:
Damage state assessment method for pinch roll based on multi-domain information deep fusion
Author(s):
XU Zeng-bing;  HUANG Zheng
(Key Laboratory of Metallurgical Equipment and Control Technology of Ministry of Education,Wuhan University of Science and Technology, Wuhan 430081, Hubei, China)
Keywords:
mechanical engineering;  damage state diagnosis;  deep learning;  signal processing;  weighted soft voting
PACS:
TH113
DOI:
10.19721/j.cnki.1671-8879.2026.01.014
Abstract:
To address the insufficient characterization capability of single-time-domain vibration signals and the inadequate diagnostic performance of single deep learning models, which result in limited diagnostic accuracy for the damage state identification of pinch rolls, a pinch roll damage state assessment method integrating multi-domain information was proposed.To fully utilize the characteristics of signals from each domain, the original vibration signal, the frequency-domain signal based on fast Fourier transform(FFT), and the time-frequency diagram based on continuous wavelet transform(CWT)were input into the Yu norm-based deep metric learning model(Yu_DML), the deep belief network(DBN), and the AlexNet convolutional neural network for preliminary diagnostic analysis, respectively. Then, combining a decision-level fusion strategy based on weighted soft voting, the recognition performance of each deep learning model and the advantages of complementary features from multi-domain signals were fully leveraged to obtain the final diagnostic result. To verify the effectiveness of the proposed method, vibration signal data from a pinch roll device in a steel plant were collected and a damage assessment experiment was conducted. The research results demonstrate that the proposed multi-domain information fusion method significantly improves diagnostic accuracy. Compared with the single Yu_DML, DBN, and AlexNet models, the diagnostic accuracy of the proposed method is improved by 28%, 2.8%, and 4.0%, respectively, which confirms the effectiveness of integrating multi-domain signals and multiple models. Through comparative experiments with fusion methods based on simple soft voting and simple hard voting, the weighted soft voting strategy adopted in this method improves diagnostic accuracy by 2.0% and 2.8%, respectively. This research result highlights that the weighted soft voting fusion strategy reasonably allocates weights according to the recognition rates of different deep learning models, thereby integrating diagnostic information from each base model. Additionally, noise resistance experiments confirm that the proposed method possesses certain generalization capabilities and has potential for transfer applications.3 tabs, 12 figs, 19 refs.

References:

[1] 史静娴,刘 娜.热连轧机卷取夹送辊位置和压力控制[J].数字技术与应用,2010(2):98-99.
SHI Jing-xian, LIU Na. Position and pressure control of pinch roll in hot continuous rolling mill coiler[J]. Digital Technology & Application, 2010(2): 98-99.
[2]李 文,赵 磊,刘 宁,等.夹送辊在热轧带钢卷取中的应用及改进[J].四川冶金,2014,36(5):19-22.
LI Wen, ZHAO Lei, LIU Ning, et al. Application and improvement of pinch rolls in hot-rolled strip coiling[J]. Sichuan Metallurgy, 2014, 36(5): 19-22.
[3]马克允.高速棒材夹送辊设备故障分析及技术改造[J].福建冶金,2018,47(5):31-33.
MA Ke-yun. Fault analysis and technical modification of high-speed bar pinch roll equipment[J]. Fujian Metallurgy, 2018, 47(5): 31-33.
[4]豆 旺,冉启武,罗德虎,等.基于改进GA优化DBN的变压器故障诊断[J/OL]. 武汉大学学报(工学版),2023:1-8.https://link.cnki.net/urlid/42.1675.T.20230926.1322.002.
DOU Wang, RAN QI-wu, LUO De-hu, et al. Transformer fault diagnosis based on improved GA optimized DBN[J/OL]. Engineering Journal of Wuhan University, 2023: 1-8.https://link.cnki.net/urlid/42.1675.T.20230926.1322.002.
[5]陈 鲲,茆志伟,张进杰,等.基于和声搜索优化栈式自编码器的柴油发动机故障诊断[J].机械工程学报,2020,56(11):132-140.
CHEN Kun, MAO Zhi-wei, ZHANG Jin-jie, et al. Diesel engine fault diagnosis based on harmony search optimized stacked autoencoder[J]. Journal of Mechanical Engineering, 2020, 56(11): 132-140.
[6]李震球.利用改进CNN实现轴承自适应抗噪故障诊断[J].机械设计与制造,2024(8):282-286,293.
LI Zhen-qiu. Adaptive anti-noise bearing fault diagnosis using improved CNN[J]. Machinery Design & Manufacture, 2024(8): 282-286, 293.
[7]王 磊.斜盘式柱塞泵的故障仿真分析与状态检测系统设计[D].石家庄:石家庄铁道大学,2024.
WANG Lei. Fault simulation analysis and condition monitoring system design for swashplate axial piston pump[D]. Shijiazhuang: Shijiazhuang Tiedao University, 2024.
[8]ZHAO M H, KANG M, TANG B P, et al. Deep residual networks with dynamically weighted wavelet coefficients for fault diagnosis of planetary gearboxes[J]. IEEE Transactions on Industrial Electronics, 2018, 65(5): 4290-4300.
[9]陈国成,张 建,菅光雷.基于改进叠加自动编码器轴承智能故障诊断方法[J].噪声与振动控制,2022,42(1):156-161.
CHEN Guo-cheng, ZHANG Jian, JIAN Guang-lei. Intelligent bearing fault diagnosis method based on improved stacked autoencoder[J]. Noise and Vibration Control, 2022, 42(1): 156-161.
[10]APPANA D K, ALEXANDER P, JONG-MYON K. Reliable fault diagnosis of bearings with varying rotational speeds using envelope spectrum and convolution neural networks[J]. Soft Computing, 2018, 22: 6719-6729.
[11]樊翔翔,项载毓,孙瑞雪,等.基于小波时频分析和Inception-BiGRU模型的盾构滚刀偏磨故障诊断[J].振动与冲击,2023,42(15):232-240.
FAN Xiang-xiang, XIANG Zai-yu, SUN Rui-xue, et al. Shield cutterhead eccentric wear fault diagnosis based on wavelet time-frequency analysis and Inception-BiGRU model[J]. Journal of Vibration and Shock, 2023, 42(15): 232-240.
[12]LUUKKA P.Similarity classifier using similarity measure derived from Yu's norms in classification of medical Datasets[J]. Computers in Biology and Medicine, 2007, 37(8): 1133-1140.
[13]YAN S, XU D, ZHANG B, et al. Graph embedding: A general framework for dimensionality reduction[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2007, 29(1): 40.
[14]李益兵,黄定洪,马建波,等.基于深度置信网络与信息融合的齿轮故障诊断方法[J].振动与冲击,2021,40(8):62-69.
LI Yi-bing, HUANG Ding-hong, MA Jian-bo, et al. Gear fault diagnosis method based on deep belief network and information fusion[J]. Journal of Vibration and Shock, 2021, 40(8): 62-69.
[15]聂 昕,谭 天,申丹凤.基于深度学习的汽车梁类件冲压回弹研究[J].中国机械工程,2023,34(7):838-846.
NIE Xin, TAN Tian, SHEN Dan-feng. Research on stamping spring back of automotive beam parts based on deep learning[J]. China Mechanical Engineering, 2023, 34(7): 838-846.
[16]向小民,汪 杰,卢 云.基于小波时频图和AlexNet网络的电弧故障诊断方法[J].国外电子测量技术,2022,41(10):170-177.
XIANG Xiao-min, WANG Jie, LU Yun. Arc fault diagnosis method based on wavelet time-frequency diagram and AlexNet network[J]. Foreign Electronic Measurement Technology, 2022, 41(10): 170-177.
[17]张 桦,沈 菲,蒋世豪,等.集成加权软投票的众包真值推理方法[J].清华大学学报(自然科学版),2022,62(2):347-354.
ZHANG Hua, SHEN Fei, JIANG Shi-hao, et al. Crowdsourcing truth inference method integrating weighted soft voting[J]. Journal of Tsinghua University(Science and Technology), 2022, 62(2): 347-354.
[18]卓德兵,曹 晖.基于小波时频图与轻量级卷积神经网络的螺栓连接损伤识别[J].工程力学,2021,38(9):228-238.
ZHUO De-bing, CAO Hui. Bolt connection damage identification based on wavelet time-frequency diagram and lightweight convolutional neural network[J]. Engineering Mechanics, 2021, 38(9): 228-238.
[19]李小娟,徐增丙,熊 文,等.基于深度度量学习的轴承故障诊断方法[J].振动与冲击,2020,39(15):25-31.
LI Xiao-juan, XU Zeng-bing, XIONG Wen, et al. Bearing fault diagnosis method based on deep metric learning[J]. Journal of Vibration and Shock, 2020, 39(15): 25-31.

Memo

Memo:
-
Last Update: 2026-02-20