基于分级插值与高斯过程回归代理模型的小样本电缆寿命预测

doi:10.3901/JME.260463

机械工程学报 ›› 2026, Vol. 62 ›› Issue (14): 198-206.doi: 10.3901/JME.260463

扫码分享

基于分级插值与高斯过程回归代理模型的小样本电缆寿命预测

王匀¹, 李政铭¹, 李文涛², 郑学文², 沈国涛¹

1. 江苏大学机械工程学院镇江 212013;
2. 磐翼信息科技(上海)有限公司上海 201100

收稿日期:2025-08-01 修回日期:2026-01-05 发布日期:2026-08-29
作者简介:王匀(通信作者)，男，1975年出生，博士，教授，博士研究生导师。主要研究方向为重大装备、结构优化、塑性成形技术、模具、微成形及控制、激光成形新技术。E-mail：wangyun@ujs.edu.cn;李政铭，男，2000年出生，硕士研究生。主要研究方向为电缆疲劳。E-mail：805650552@qq.com;李文涛，男，1989年出生，硕士，工程师。主要研究方向为柔性管线运动仿真。E-mail：liwentao@pan-i.com;郑学文，男，1999年出生，硕士，工程师。主要研究方向为柔性管线运动仿真。E-mail：zhengxuewen@pan-i.com;沈国涛，男，2001年出生，硕士。主要研究方向为自主移动机器人。E-mail：3463810448@qq.com
基金资助:
泰州市科技支撑计划(工业)竞争TG202409、镇江市重点研发计划(GY2023013)和中山市社会公益与基础研究专项(2024B2046)资助项目。

Gaussian Process Regression-based Surrogate Model for Small-sample Cable Fatigue Life Prediction under Stratified Interpolation

WANG Yun¹, LI Zhengming¹, LI Wentao², ZHENG Xuewen², SHEN Guotao¹

1. School of Mechanical Engineering, Jiangsu University, Zhenjiang 212013;
2. PanYi Information Technology (Shanghai) Co., Ltd., Shanghai 201100

Received:2025-08-01 Revised:2026-01-05 Published:2026-08-29

摘要/Abstract

摘要： 电缆在高低温环境下和动态弯曲载荷条件下的疲劳寿命预测是汽车工程中的关键挑战。传统试验方法存在耗时长、成本高的问题，且材料试验后无法复测；数值仿真只能给出阈值范畴，而且对计算资源要求高。针对上述问题，本研究构建了融合物理知识与数据驱动的高斯过程回归代理模型(Gaussian process regression-based surrogate model,GPRSM)，实现了电缆疲劳寿命的高效预测。首先基于设计分级插值策略，结合温度-寿命衰减物理约束，使模型能够覆盖全域工况，保证数据合理性和完备性；提出了小样本学习方法通过弯曲半径、温度等多维关键参数的物理关联性分析，实现小样本场景下的电缆疲劳寿命高精度预测；利用GPRSM的不确定性量化方法，输出寿命预测均值与置信区间，为汽车安全性和电缆可靠性设计提供定量依据。据研究表明，该方法在全工况范围内预测性能良好，高疲劳寿命区预测值偏保守，适用于注重安全裕度的工程应用场景，整体决定系数(R²)达到0.86。本研究为电缆疲劳寿命的预测提供了高效可靠的技术手段，具有重要的工程应用价值和推广意义。

关键词: 温度-载荷耦合约束, 电缆疲劳预测, 高斯过程回归代理模型, 分级插值策略, 小样本学习方法

Abstract: Predicting the fatigue life of cable under extreme temperature variations and dynamic bending loads is a critical challenge in automotive engineering. Conventional experimental testing is time-consuming, costly, and inherently destructive, precluding retesting on the same specimen; while numerical simulations often yield only qualitative or threshold-level results and demand substantial computational resources. To address these limitations, this study proposes a Gaussian process regression-based surrogate model (GPRSM) that integrates physical knowledge with data-driven learning for efficient and accurate fatigue life prediction. A stratified interpolation strategy is first designed to augment sparse experimental data, incorporating physically informed constraints on temperature-dependent life degradation to ensure data completeness and physical consistency across the full operational envelope. A small-sample learning framework is then developed by analyzing the physical correlations among multidimensional parameters—such as bending radius and temperature—enabling high-precision predictions under limited-sample conditions. Leveraging the inherent uncertainty quantification capability of GPRSM, the model outputs both the mean predicted fatigue life and corresponding confidence intervals, providing a quantitative foundation for automotive safety assessment and reliability-driven cable design. Demonstrating an overall R² of 0.86, the method delivers reliable predictions across all operating conditions and conservative estimates at high fatigue lives, aligning well with safety-critical engineering needs. This work presents an efficient and robust technical solution for cable fatigue life prediction, offering significant value for engineering applications and potential for broader adoption in reliability analysis of complex electromechanical systems.

Key words: temperature-load coupling constraints, cable fatigue prediction, Gaussian process regression-based surrogate model (GPRSM), stratified interpolation strategy, small-sample learning method

中图分类号:

TH17

王匀, 李政铭, 李文涛, 郑学文, 沈国涛. 基于分级插值与高斯过程回归代理模型的小样本电缆寿命预测[J]. 机械工程学报, 2026, 62(14): 198-206.

WANG Yun, LI Zhengming, LI Wentao, ZHENG Xuewen, SHEN Guotao. Gaussian Process Regression-based Surrogate Model for Small-sample Cable Fatigue Life Prediction under Stratified Interpolation[J]. Journal of Mechanical Engineering, 2026, 62(14): 198-206.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.260463

http://www.cjmenet.com.cn/CN/Y2026/V62/I14/198

参考文献

[1] CRUZADO A，LUCARNIN S，LORCA J，et al. Microstructure-based fatigue life model of metallic alloys with bilinear Coffin-Manson behavior[J]. International Journal of Fatigue，2018，107：40-48.
[2] BAI Y，ZHAO J. A novel transformer-based multi-variable multi-step prediction method for chemical process fault prognosis[J]. Process Safety and Environmental Protection，2023，169：937-947.
[3] HUANG Z. Research on multi variable time series fault prediction of gas turbine based on transformer model[C]//2024 International Seminar on Artificial Intelligence，Computer Technology and Control Engineering (ACTCE). IEEE，2024：546-551.
[4] WANG H，LI B，GONG J，et al. Machine learning-based fatigue life prediction of metal materials：Perspectives of physics-informed and data-driven hybrid methods[J]. Engineering Fracture Mechanics，2023，284：109242.
[5] HE G Y，ZHAO Y X，YAN C L. Uncertainty quantification in multiaxial fatigue life prediction using Bayesian neural networks[J]. Engineering Fracture Mechanics，2024，298：109961.
[6] LOUTAS T，ELEFTHEROGLOU N，ZAROUCHARS D. A data-driven probabilistic framework towards the in-situ prognostics of fatigue life of composites based on acoustic emission data[J]. Composite Structures，2017，161：522-529.
[7] FARID M. Data-driven method for real-time prediction and uncertainty quantification of fatigue failure under stochastic loading using artificial neural networks and Gaussian process regression[J]. International Journal of Fatigue，2022，155：106415.
[8] TAKI M，ROHANI A，SOHEILI-FARD F，et al. Assessment of energy consumption and modeling of output energy for wheat production by neural network (MLP and RBF) and Gaussian process regression (GPR) models[J]. Journal of Cleaner Production，2018，172：3028-3041.
[9] JAMEI M，AHMADIANFAR I，OLUMEGBON I A，et al. On the assessment of specific heat capacity of nanofluids for solar energy applications：Application of Gaussian process regression (GPR) approach[J]. Journal of Energy Storage，2021，33：102067.
[10] KONG D，CHEN Y，LI N. Gaussian process regression for tool wear prediction[J]. Mechanical Systems and Signal Processing，2018，104：556-574.
[11] 刘志状，吴昊. 一种基于参数影响的数据驱动下的疲劳寿命预测方法[J]. 机械工程学报，2023，59(4)：71-79. LIU Zhizhuang，WU Hao. Data-driven fatigue life prediction method based on the influence of parameters[J]. Journal of Mechanical Engineering，2023，59(4)：71-79.
[12] ZHOU X，CHEN Z，GU C，et al. Thermo-mechanical coupling analysis of the end faces for a mechanical seal under dry friction[J]. Tribology International，2021，160：107050.
[13] XU J，YANG H，LI H，et al. Significance-based optimization of processing parameters for thin-walled aluminum alloy tube NC bending with small bending radius[J]. Transactions of Nonferrous Metals Society of China，2012，22(1)：147-156.
[14] ZHOU X，LIANG W，SHIMIZU S，et al. Siamese neural network based few-shot learning for anomaly detection in industrial cyber-physical systems[J]. IEEE Transactions on Industrial Informatics，2020，17(8)：5790-5798.
[15] DE Beaulieu M H，JHA M S，GARNIER H，et al. Remaining useful life prediction based on physics- informed data augmentation[J]. Reliability Engineering & System Safety，2024，252：110451.
[16] ZHANG J，LI J，PEI W，et al. Physics-informed data augmentation of experimental erosion aata through generative adversarial networks[C]//Fluids Engineering Division Summer Meeting. American Society of Mechanical Engineers，2024，88131：V002T05A013.
[17] SKALA V. RBF interpolation with CSRBF of large data sets[J]. Procedia Computer Science，2017，108：2433-2437.
[18] CUOMO S，GALLETTI A，GIUNTA G，et al. Reconstruction of implicit curves and surfaces via RBF interpolation[J]. Applied Numerical Mathematics，2017，116：157-171.
[19] LAMBAY A，LIU Y，MORGAN P，et al. Generating synthetic datasets with deep learning models for human physical fatigue analysis[J]. Machines，2025，13(3)：235.
[20] LACASA M，PRADOS F，ALEGRE J，et al. A synthetic data generation system for myalgic encephalomyelitis/ chronic fatigue syndrome questionnaires[J]. Scientific Reports，2023，13(1)：14256.
[21] CHEN Y，CHEN X，WANG Y，et al. The interpolation of sparse geophysical data[J]. Surveys in Geophysics，2019，40：73-105.
[22] TRAD D. Five-dimensional interpolation：Recovering from acquisition constraints[J]. Geophysics，2009，74(6)：V123-V132.
[23] KOHOUT J. Modified Arrhenius equation in materials science，chemistry and biology[J]. Molecules，2021，26(23)：7162.

基于分级插值与高斯过程回归代理模型的小样本电缆寿命预测

Gaussian Process Regression-based Surrogate Model for Small-sample Cable Fatigue Life Prediction under Stratified Interpolation

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	许伟洋, 杨兆军, 刘志峰, 何佳龙, 李国发, 陈传海, 刘严. 基于动态多尺度注意力机制的数控机床伺服刀架动力头退化状态轻量级识别方法[J]. 机械工程学报, 2026, 62(10): 348-361.
[2]	杨彬, 李雅宁, 雷亚国, 李响, 曹军义, 武通海. 迁移拓扑规划的机械设备群体协同智能诊断方法[J]. 机械工程学报, 2026, 62(4): 1-11.
[3]	周坪, 周公博, 王晗宇, 王攀, 闫晓东, 李远博. 基于涡流热成像的钢丝绳表面缺陷红外视觉检测方法[J]. 机械工程学报, 2026, 62(4): 86-97.
[4]	黄金凤, 王成城, 何宏亮, 王旭, 李奇, 杨康定, 王凯, 张飞斌, 秦朝烨, 褚福磊. 思维链范式智能运维文本基多模态智能体[J]. 机械工程学报, 2025, 61(23): 58-74.
[5]	杨波, 申小玉, 王时龙, 何彦, 杜卡泽. 面向设备运维的人机物三元融合知识图谱构建方法[J]. 机械工程学报, 2025, 61(17): 215-232.
[6]	肖扬, 王华庆, 李华, 王庆锋. 基于精细复合缩放多尺度加权排列熵的跨域故障诊断方法[J]. 机械工程学报, 2025, 61(16): 28-39.
[7]	孙世博, 袁静, 赵倩, 蒋会明, 魏颖. 内核噪声拓展自适应多元变分模态分解及机械复合故障诊断应用[J]. 机械工程学报, 2025, 61(11): 171-182.
[8]	郑俊康, 王辉, 向家伟. 动力学模型驱动的RV减速器故障诊断方法[J]. 机械工程学报, 2025, 61(11): 194-207.
[9]	李永杰, 张周锁, 罗欣. 带参考信号的频域盲解卷积算法及其在卫星微振动同频相关源定量辨识中的应用[J]. 机械工程学报, 2025, 61(10): 215-229.
[10]	邵海东, 颜深, 刘政武, 肖一鸣, 韩特. 面向小样本场景的风机多增量故障诊断[J]. 机械工程学报, 2025, 61(10): 230-240.
[11]	吴艳灵, 汤宝平, 邓蕾, 付豪. 低通筛选优化神经架构搜索的风电齿轮箱边缘侧故障诊断方法[J]. 机械工程学报, 2025, 61(7): 361-372.
[12]	雷亚国, 李熹伟, 李响, 李乃鹏, 杨彬. 面向机械设备通用健康管理的智能运维大模型[J]. 机械工程学报, 2025, 61(6): 1-13.
[13]	鲍宏, 王政, 陶璟, 李红真, 于随然, 宋培龙. 面向重用与装配复杂度均衡性的机电产品非合作博弈模块化设计方法[J]. 机械工程学报, 2025, 61(5): 285-296.
[14]	李华, 王天杨, 张飞斌, 褚福磊. KurVMDPgram：一种用于旋转机械故障诊断的信号分解算法[J]. 机械工程学报, 2025, 61(4): 11-23.
[15]	裴雪武, 李新宇, 高亮, 高艺平, 陈志敏. 复合故障退化情形下基于自适应非线性状态估计的装备状态监测健康指标[J]. 机械工程学报, 2025, 61(3): 119-129.