数据与模型融合迁移的再制造产品可靠性动态预测方法

doi:10.3969/j.issn.1004-132X.2026.04.025

摘要/Abstract

摘要：

针对再制造产品可靠性因数据样本少而导致其服役过程中可靠性状态难以准确预测的问题，提出了融合相似产品物-场退化数据与模型迁移微调的再制造产品可靠性动态预测方法。首先，利用物-场模型分析影响产品可靠性的“物性”与“场性”退化指标，利用线性回归模型构建反映产品多维物-场退化特征的综合退化指标，并设计一个三阶段相似性计算方法来筛选和迁移相似产品历史退化数据以扩充样本。然后，针对相似产品历史物-场退化数据的空间耦合性和时间依赖性特征，建立基于卷积长短期神经网络的可靠性预测模型。进一步，通过深度迁移技术动态调整预测模型参数，提高模型对再制造产品在个性化服役场景下的可靠性预测精度。最后，以某再制造主轴系统为例对所提预测方法进行验证，预测结果的决定系数（R²）达到0.92，表明了该方法的有效性。

关键词: 再制造产品, 可靠性预测, 性能退化, 相似性计算, 深度迁移学习

Abstract:

To address the problems that the reliability data samples of remanufactured products were scarce， leading to difficulties in accurately predicting their reliability status during service， a dynamic reliability prediction method for remanufactured products was proposed， integrating substance-field degradation data from similar products with model transfer fine-tuning. Firstly， the “physical form” and “field properties” degradation indicators affecting product reliability were analyzed using the substance-field model. Then， a comprehensive degradation index reflecting the multi-dimensional substance-field degradation characteristics of products was constructed using a linear regression model， and a three-stage similarity calculation method was designed to screen and transfer historical degradation data from similar products for sample expansion. Secondly， to address the spatial coupling and temporal dependency characteristics of the historical substance-field degradation data of similar products， a reliability prediction model was established based on a convolutional long short-term memory neural network. Furthermore， the parameters of the prediction model were dynamically adjusted through deep transfer learning techniques to improve the prediction accuracy for the reliability of remanufactured products under personalized service scenarios. Finally， the proposed prediction method was validated using a remanufactured spindle system as a case study， and the coefficient of determination （R²） of the prediction results reache 0.92， which indicates the effectiveness of the method.

Key words: remanufactured product, reliability prediction, performance degradation, similarity calculation, deep transfer learning

中图分类号:

TH122

冯雨康, 朱硕, 江志刚, 鄢威, 张华. 数据与模型融合迁移的再制造产品可靠性动态预测方法[J]. 中国机械工程, 2026, 37(4): 1007-1015.

FENG Yukang, ZHU Shuo, JIANG Zhigang, YAN Wei, ZHANG Hua. Reliability Dynamic Prediction Method for Remanufactured Products Based on Data-model Integration and Transfer[J]. China Mechanical Engineering, 2026, 37(4): 1007-1015.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.cmemo.org.cn/CN/10.3969/j.issn.1004-132X.2026.04.025

https://www.cmemo.org.cn/CN/Y2026/V37/I4/1007

图/表 10

图1 再制造产品服役阶段可靠性预测流程

Fig.1 Process for predicting the reliability of remanufactured products during service period

图2 基于功能分解和物-场模型的产品退化指标分析

Fig.2 Analysis of product degradation indicators via functional decomposition and substance-field model

图3 时间窗处理图示

Fig.3 Illustration of time window processing

图4 可靠性预测模型微调方法框架

Fig.4 Framework of fine-tuning method for reliability prediction model

表1 主轴系统物-场退化数据类型

Tab.1 Data types of substance-field degradation of the spindle system

项目	退化指标名称	符号
物	主轴形变量/μm	S₁
	主轴裂纹深度/μm	S₂
	轴承磨损深度/μm	S₃
	轴承径向间隙量/μm	S₄
	油膜厚度/μm	S₅
	油箱密封泄漏量/（mL·min $- 1$ ）	S₆
场	主轴径向跳动量/μm	F₁
	功率损耗率/%	F₂
	轴承温升比率/%	F₃
	噪声/dB	F₄
	振动/（mm·s $- 1$ ）	F₅

表1 主轴系统物-场退化数据类型

Tab.1 Data types of substance-field degradation of the spindle system

项目	退化指标名称	符号
物	主轴形变量/μm	S₁
	主轴裂纹深度/μm	S₂
	轴承磨损深度/μm	S₃
	轴承径向间隙量/μm	S₄
	油膜厚度/μm	S₅
	油箱密封泄漏量/（mL·min $- 1$ ）	S₆
场	主轴径向跳动量/μm	F₁
	功率损耗率/%	F₂
	轴承温升比率/%	F₃
	噪声/dB	F₄
	振动/（mm·s $- 1$ ）	F₅

表2 训练集和测试集数据细节

Tab.2 Details of training set and test set data

	训练集	测试集1	测试集2
样本数量	3	1	1
运行周期	从运行到失效	0~150	150~200
服役生命周期	1	—	—

图5 物-场退化数据可视化

Fig.5 Visualization of substance-field degradation data

表3 CNN-LSTM网络结构和训练参数设置

Tab.3 Network structure and training parameter settings of CNN-LSTM

网络设置	设置结果
卷积层数	1
池化层数	1
LSTM层数	2
全连接层数	2
卷积核数量	16
卷积核尺寸	2×2
池化窗口大小	1×1
LSTM-1神经元个数	32
LSTM-2神经元个数	16
全连接层-1神经元个数	8
全连接层-2神经元个数	4
时间窗大小	5
迭代次数	200
批量大小	8
学习率	0.01

图6 测试集2微调前后可靠性曲线预测结果

Fig.6 Reliability curve prediction results before and after fine-tuning of test set 2

表4 不同预测模型的性能比较

Tab.4 Performance comparison of different prediction models

模型	σ_RMSE	R²
CNN	0.010	0.68
CNN_TL	0.007	0.83
LSTM	0.009	0.76
LSTM_TL	0.006	0.86
CNN-LSTM_TL	0.004	0.92

参考文献 21

[1]	ZHANG Xugang， AO Xiuyi， JIANG Zhigang， et al. A Remanufacturing Cost Prediction Model of Used Parts Considering Failure Characteristics［J］. Robotics and Computer-integrated Manufacturing， 2019， 59： 291-296.
[2]	JIANG Zhigang， ZHOU Tingting， ZHANG Hua， et al. Reliability and Cost Optimization for Remanufacturing Process Planning［J］. Journal of Cleaner Production， 2016， 135： 1602-1610.
[3]	曹华军，杜彦斌. 机床装备在役再制造的内涵及技术体系［J］. 中国机械工程， 2018， 29（19）： 2357-2363.
	CAO Huajun， DU Yanbin. Connotation and Technology System Framework of In-service Remanufacturing of Machine Tools［J］. China Mechanical Engineering， 2018， 29（19）： 2357-2363.
[4]	张孟，李传军，苗百然，等. 基于Bootstrap改进的机床主轴可靠性建模［J］. 计算机集成制造系统， 2025， 31（9）： 3401-3410.
	ZHANG Meng， LI Chuanjun， MIAO Bairan， et al. Reliability Modeling of Machine Tool Spindle Based on Bootstrap Improvement［J］. Computer Integrated Manufacturing Systems， 2025， 31（9）： 3401-3410.
[5]	BOUJARIF A， COIT D W， JOUINI O， et al. A Deep-learning-based Framework to Predict the Reliability of Multicomponent Repairable Systems in a Closed-loop Supply Chain［J］. IEEE Transactions on Reliability， 2025， 74（3）： 3809-3823.
[6]	LI Chen， LIU Rongxing， PAN Fukui. Simulation of Reliability Prediction Based on Multiple Factors for Spinning Machine［J］. Autex Research Journal， 2020， 20（1）： 17-23.
[7]	LIU Zheng， LIU Xin， HUANG Hongzhong， et al. A New Inherent Reliability Modeling and Analysis Method Based on Imprecise Dirichlet Model for Machine Tool Spindle［J］. Annals of Operations Research， 2022， 311（1）： 295-310.
[8]	LI He， DENG Zhiming， GOLILARZ N A， et al. Reliability Analysis of the Main Drive System of a CNC Machine Tool Including Early Failures［J］. Reliability Engineering & System Safety， 2021， 215： 107846.
[9]	张辉，张华，江志刚，等. 基于条件分布的AMSAA模型再制造机床可靠性增长预测［J］. 机械设计与制造， 2017（6）： 127-130.
	ZHANG Hui， ZHANG Hua， JIANG Zhigang， et al. Remanufacturing Machine Reliability Growth Prediction of Army Materiel System Analysis Activity Model Based on Conditional Distribution［J］. Machinery Design & Manufacture， 2017（6）： 127-130.
[10]	LIU C H， LI W Y， RAO W Z， et al. Study on the Failure Mechanism of the Polymorphic Mixture for Remanufactured Machinery Parts［J］. Strength of Materials， 2018， 50（1）： 151-156.
[11]	XIE Chuan， ZHANG Peng， YAN Zhi. Correlation Analysis of Aeroengine Operation Monitoring Using Deep Learning［J］. Soft Computing， 2021， 25（1）： 551-562.
[12]	PAN Weihuang， FENG Yunwen， LU Cheng， et al. Analyzing the Operation Reliability of Aeroengine Using Quick Access Recorder Flight Data［J］. Reliability Engineering and System Safety， 2023， 235： 109193.
[13]	ZHU Linbo， CHEN Dong， FENG Pengfei. Equipment Operational Reliability Evaluation Method Based on RVM and PCA-fused Features［J］. Mathematical Problems in Engineering， 2021， 2021（1）： 6687248.
[14]	MATUSZCZAK M， ŻBIKOWSKI M， TEODORCZYK A. Predictive Modelling of Turbofan Engine Components Condition Using Machine and Deep Learning Methods［J］. Eksploatacja i Niezawodność-Maintenance and Reliability， 2021， 23（2）： 359-370.
[15]	REN Guoai， WANG Zhihai， LIU Xiaoqin， et al. Remaining Useful Life Prediction of Industrial Robot RV Reducer with Multiple Deep Networks and Multicore Support Vector Data Description［J］. Journal of Mechanical Science and Technology， 2024， 38（8）： 4037-4051.
[16]	文娟，宋洋，林苏奔，等．基于多域退化特征生成的RV减速器寿命预测［J/OL］．计算机集成制造系统.
	WEN Juan， SONG Yang， LIN Suben， et al. Life Prediction of RV Reducer Based on Multi-domain Degradation Feature Generation［J/OL］. Computer Integrated Manufacturing Systems..
[17]	高玉霞，王向华，王静远，等. 基于性能退化指标的轴承剩余寿命预测及其应用［J］. 兵工自动化， 2023， 42（5）： 40-45.
	GAO Yuxia， WANG Xianghua， WANG Jingyuan， et al. Based Performance Degradation Indicator RUL Prediction and Its Application in Bearing［J］. Ordnance Industry Automation， 2023， 42（5）： 40-45.
[18]	高淑芝，陈国庆，张义民，等. 基于特征选择和ELM神经网络的轴承可靠性预测［J］. 机械设计与制造， 2024（8）： 170-173.
	GAO Shuzhi， CHEN Guoqing， ZHANG Yimin， et al. Reliability Prediction of Bearing Based on Feature Selection and ELM Neural Network Network［J］. Machinery Design & Manufacture， 2024（8）： 170-173.
[19]	GUO Xin， LIU Ying， ZHAO Wu， et al. Supporting Resilient Conceptual Design Using Functional Decomposition and Conflict Resolution［J］. Advanced Engineering Informatics， 2021， 48： 101262.
[20]	许鉴，江志刚，朱硕，等. 基于废旧机电产品多属性寿命情景的再制造策略决策方法［J］. 计算机集成制造系统， 2026， 32（1）： 287-299.
	XU Jian， JIANG Zhigang， ZHU Shuo， et al. Decision-making Method for Selecting Remanufacturing Strategies Based on Multi-attribute Remaining Life Scenarios of Used Electromechanical Products［J］. Computer Integrated Manufacturing Systems， 2026， 32（1）： 287-299.
[21]	SHINIKE K. A Two Phase Method for Determining the Number of Neurons in the Hidden Layer of a 3-layer Neural Network［C］∥Proceedings of SICE Annual Conference 2010. Taipei， IEEE， 2010： 238-242.