Digital Twin-Based Life Prediction of Full-Ceramic Rolling Bearings
-
摘要:
为解决全陶瓷滚动轴承健康检测中裂纹扩展机理难以精确表征，以及极端工况下监测数据样本稀缺导致寿命预测精度受限的问题，本文提出一种可在小样本条件下实现高精度寿命预测的混合网络模型. 首先，基于轴承运行特性构建二自由度动力学模型，通过振动响应行为分析描述轴承的非线性动力学特性，并融合时域统计特征与频域能量特征开展状态信息联合表征；其次，针对脆性材料裂纹扩展过程难以直接观测的特点，将裂纹稳态扩展的物理约束引入物理信息神经网络训练中，使仿真信号演化规律与裂纹真实扩展趋势保持一致；同时，利用对抗生成神经网络建立仿真信号与真实信号的映射关系，生成具有高保真特征的模拟振动数据以扩充训练样本空间以解决小样本问题；最后，采用随机森林回归模型开展寿命预测与裂纹失稳临界状态识别，构建面向陶瓷轴承剩余寿命评估的完整分析流程. 结果表明：该方法能够准确反映裂纹从萌生至扩展阶段的动力学变化规律，寿命预测平均准确率达到99.5%以上；经物理约束融合后的仿真信号与真实信号在主要特征频带能量分布和包络谱峰值位置上呈现高一致性，裂纹扩展趋势识别能力得到增强，并可实现裂纹由稳态扩展向失稳阶段转变关键节点的辨识.
Abstract:To address two challenges in the health monitoring of full-ceramic rolling bearings, namely the difficulty in accurately characterizing the crack propagation mechanism and the limited accuracy of life prediction caused by the scarcity of monitoring data samples under extreme operating conditions, a hybrid network model for high-precision life prediction under small-sample conditions was developed. A two-degree-of-freedom dynamic model was established based on bearing operating characteristics to analyze and describe the nonlinear dynamic behavior of the bearing through vibration responses. Time-domain statistical and frequency-domain energy features were combined to conduct a joint characterization of state information. By considering the characteristic that the crack propagation process of brittle materials is difficult to directly observe, physical constraints governing stable crack propagation were integrated into the physics-informed neural network training, ensuring simulated signal evolution aligned with actual crack propagation trends. A generative adversarial neural network mapped simulated signals to real signals, producing high-fidelity simulated vibration data to expand the training sample space and alleviate small-sample limitations. A random forest regression model was applied for life prediction and identification of critical crack instability states, forming a complete analysis process of residual life assessment for ceramic bearings. Results indicate that the method accurately captures the laws of dynamic evolution from crack initiation to propagation, achieving an average life prediction accuracy of over 99.5%. The simulated signals fused with physical constraints show high consistency with real signals in the energy distribution of key feature frequency bands and peak locations of the envelope spectrum. This method improves the capability of identifying crack propagation trends and enables the recognition of the critical node corresponding to the transition from stable to unstable crack propagation.
-
Key words:
- full-ceramic rolling bearing /
- small-sample /
- crack propagation /
- life prediction
-
图 2 陶瓷球与内圈滚道的实际接触状
Figure 2. Actual contact state between ceramic ball and inner raceway
图 8 训练样本数据时、频域信号图像
Figure 8. Time-domain and frequency-domain signals of training sample data
图 13 原始数据1时、频域信号
Figure 13. Time-domain and frequency-domain signals of original data 1
图 14 真实与仿真的裂纹尺寸及时、频域信号对比
Figure 14. Comparison of crack size and time-domain and frequency-domain signals between real and simulated data
图 15 原始数据2时、频域信号
Figure 15. Time-domain and frequency-domain signals of original data 2
图 19 有噪音与无噪音状态特征对比
Figure 19. Comparison of state features between noisy and noise-free conditions
表 1 数据集信息
Table 1. Dataset information
试验序号 数据长度
（文件个数）/个工况 原始数据1 123 径向载荷：4 kN
转速：4000 转/min原始数据2 12 径向载荷：5 kN
转速：4000 转/min原始数据3 271 径向载荷：4 kN
转速：6000 转/min原始数据4 100 径向载荷：4 kN
转速：6000 转/min原始数据5 7 径向载荷：5 kN
转速：5000 转/min原始数据6 26 径向载荷4.5 kN
转速：6000 转/min
下载: 导出CSV
表 2
6203 轴承物理参数Table 2. Physical parameters of
6203 bearing参数名称 数值 参数名称 数值 轴承外径/mm 40.0 节圆直径/mm 28.5 轴承内径/mm 17.0 滚动体个数 8 径向游隙/μm 10.0 接触角/（°） 0
下载: 导出CSV
表 3 氧化锆ZrO2的材料特性
Table 3. Material properties of zirconia （ZrO2）
参数名称 数值 密度/（g•cm−3） 5.70 ～ 6.05 热膨胀系数/（× 10−6K） 7.00 ～ 10.50 弹性模量/GPa 180 ～ 210 泊松比 0.30 硬度/HV 800 ～ 1500 抗弯强度 900 ～ 1200 抗压强度/MPa 1000 ～3000 断裂韧性/（MPa•m1/2） 8 ～ 10 导热系数/（W•m−1•K−1） 2 ～ 3 磁性 无 电绝缘性 绝缘
下载: 导出CSV
-
[1] Tian J X, Wu Y H, Sun J, et al. Thermal dynamic exploration of full-ceramic ball bearings under the self-lubrication condition[J]. Lubricants, 2022, 10(9): 213. doi: 10.3390/lubricants10090213 [2] Baranger E, Cluzel C, Ladevèze P, et al. Effects of the thermomechanical loading path on the lifetime prediction of self-healing ceramic matrix composites[J]. Science and Engineering of Composite Materials, 2011, 18(4): 259-264 doi: 10.1515/secm.2011.050 [3] Sun Y, Xie H, Liu L J, et al. Enhanced mechanical properties in ceramic multilayer composites through integrating crystallographic texture and second-phase toughening[J]. Ceramics International, 2021, 47(22): 31222-31228. doi: 10.1016/j.ceramint.2021.07.298 [4] 王彪, 秦勇, 贾利民, 等. 监测数据驱动的城轨列车轴箱轴承剩余寿命预测[J]. 西南交通大学学报, 2024, 59(1): 229-238. doi: 10.3969/j.issn.0258-2724.20220230Wang Biao, Qin Yong, Jia Limin, et al. Monitoring data-driven prediction of remaining useful life of axle-box bearings for urban rail transit trains[J]. Journal of Southwest Jiaotong University, 2024, 59(1): 229-238. doi: 10.3969/j.issn.0258-2724.20220230 [5] Kunzelmann B, Rycerz P, Xu Y L, et al. Prediction of rolling contact fatigue crack propagation in bearing steels using experimental crack growth data and linear elastic fracture mechanics[J]. International Journal of Fatigue, 2023, 168: 107449. doi: 10.1016/j.ijfatigue.2022.107449 [6] Khader I, Rasche S, Lube T, et al. Lifetime prediction of ceramic components–A case study on hybrid rolling contact[J]. Engineering Fracture Mechanics, 2017, 169: 292-308. doi: 10.1016/j.engfracmech.2016.10.014 [7] Wu B, Zhang X H, Shi H, et al. Failure mode division and remaining useful life prognostics of multi-indicator systems with multi-fault[J]. Reliability Engineering & System Safety, 2024, 244: 109961. doi: 10.1016/j.ress.2024.109961 [8] Salahi Nezhad M, Larsson F, Kabo E, et al. Numerical prediction of railhead rolling contact fatigue crack growth[J]. Wear, 2023, 530/531: 205003. [9] Sevener K M, Tracy J M, Chen Z, et al. Crack opening behavior in ceramic matrix composites[J]. Journal of the American Ceramic Society, 2017, 100(10): 4734-4747. doi: 10.1111/jace.14976 [10] Li J P, Zhou Y, Lu Z C, et al. Experimental study on the mechanism of wheel-rail steels crack initiation and wear growth under rolling contact fatigue[J]. Procedia Structural Integrity, 2022, 37: 582-589. doi: 10.1016/j.prostr.2022.01.126 [11] Vetrone J, Obregon J E, Indacochea E J, et al. The characterization of deformation stage of metals using acoustic emission combined with nonlinear ultrasonics[J]. Measurement, 2021, 178: 109407. doi: 10.1016/j.measurement.2021.109407 [12] Li T, Shi H T, Bai X T, et al. A digital twin model of life-cycle rolling bearing with multiscale fault evolution combined with different scale local fault extension mechanism[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 3507211. doi: 10.1109/tim.2023.3243663 [13] Li S, Jiang Q B, Xu Y D, et al. Digital twin-driven focal modulation-based convolutional network for intelligent fault diagnosis[J]. Reliability Engineering & System Safety, 2023, 240: 109590. doi: 10.1016/j.ress.2023.109590 [14] Ying J, Yang Z J, Chen C H, et al. Remaining useful life prediction of rolling element bearing based on hybrid drive of data-driven and dynamic model[J]. Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science, 2022: 09544062221141333. [15] Zhao W T, Zhang C, Fan B, et al. Research on rolling bearing virtual-real fusion life prediction with digital twin[J]. Mechanical Systems and Signal Processing, 2023, 198: 110434. doi: 10.1016/j.ymssp.2023.110434 [16] Ye Y M, Yang Q, Yang F, et al. Digital twin for the structural health management of reusable spacecraft: a case study[J]. Engineering Fracture Mechanics, 2020, 234: 107076. doi: 10.1016/j.engfracmech.2020.107076 [17] Qin Y, Wu X G, Luo J. Data-model combined driven digital twin of life-cycle rolling bearing[J]. IEEE Transactions on Industrial Informatics, 2022, 18(3): 1530-1540. doi: 10.1109/TII.2021.3089340 [18] Zhang R, Zeng Z Q, Li Y F, et al. Research on remaining useful life prediction method of rolling bearing based on digital twin[J]. Entropy, 2022, 24(11): 1578. doi: 10.3390/e24111578 [19] Lu Q B, Li M. Digital twin-driven remaining useful life prediction for rolling element bearing[J]. Machines, 2023, 11(7): 678. doi: 10.3390/machines11070678 [20] An D, Shi K, Shao M, et al. Research on full-life rolling bearing virtual-real fusion defect prediction with digital twin[J]. Engineering Research Express, 2025, 7(2): 025538. doi: 10.1088/2631-8695/adb7b9 [21] Cai Z X, Liu Y W, Zhang W, et al. A rolling bearing fault diagnosis method based on generalized dataset realizing small sample transfer learning[C]//2023 6th International Conference on Robotics, Control and Automation Engineering (RCAE). Suzhou, China. IEEE, 2023: 416-424. [22] Cui L L, Xiao Y C, Liu D D, et al. Digital twin-driven graph domain adaptation neural network for remaining useful life prediction of rolling bearing[J]. Reliability Engineering & System Safety, 2024, 245: 109991. doi: 10.1016/j.ress.2024.109991 [23] Cao W, Meng Z, Li J M, et al. A remaining useful life prediction method for rolling bearing based on TCN-transformer[J]. IEEE Transactions on Instrumentation and Measurement, 2025, 74: 3501309. [22] Cao W, Meng Z, Li J M, et al. A remaining useful life prediction method for rolling bearing based on TCN-transformer[J]. IEEE Transactions on Instrumentation and Measurement, 2025, 74: 3501309. [24] Wang Q S, Sun Z C, Zhu Y M, et al. A fault diagnosis method based on an improved diffusion model under limited sample conditions[J]. PLoS One, 2024, 19(9): e0309714. [23] Wang Q S, Sun Z C, Zhu Y M, et al. A fault diagnosis method based on an improved diffusion model under limited sample conditions[J]. PLoS One, 2024, 19(9): e0309714. [25] Cai Z X, Liu Y W, Zhang W, et al. A rolling bearing fault diagnosis method based on generalized dataset realizing small sample transfer learning[C]//2023 6th International Conference on Robotics, Control and Automation Engineering (RCAE). Suzhou, China. IEEE, 2023: 416-424. [26] 孟春成, 亐道远, 段晓晨. 城市轨道交通土建工程造价非线性预测与反演[J]. 西南交通大学学报, 2025, 60(1): 137-146. doi: 10.3969/j.issn.0258-2724.20230209Meng Chuncheng, Yu Daoyuan, Duan Xiaochen. Nonlinear prediction and inversion of civil engineering cost of urban rail transit[J]. Journal of Southwest Jiaotong University, 2025, 60(1): 137-146. doi: 10.3969/j.issn.0258-2724.20230209 [27] 易彩, 林建辉, 汪浩, 等. VMD引导的轮对与轴承复合故障诊断方法[J]. 西南交通大学学报, 2024, 59(1): 151-159. doi: 10.3969/j.issn.0258-2724.20211088Yi Cai, Lin Jianhui, Wang Hao, et al. Compound fault diagnosis method guided by variational mode decomposition for wheelsets and bearings[J]. Journal of Southwest Jiaotong University, 2024, 59(1): 151-159. doi: 10.3969/j.issn.0258-2724.20211088 -
下载:
点击查看大图
计量
- 文章访问数: 23
- HTML全文浏览量: 10
- PDF下载量: 4
- 被引次数: 0