Data-driven prediction and multi-parameter optimization of fatigue life of commercial vehicle drive shafts

doi:10.3969/j.issn.1674-8484.2026.02.005

Abstract

Abstract:

A data-driven method was proposed to efficiently and accurately predict and optimize the fatigue life of vehicle drive shaft assemblies. Based on experimental testing and finite element analysis (FEA), a deep neural network (DNN) was utilized to optimize the parameters of key components. The reliability of the finite element model for the drive shaft assembly was verified through strain testing, based on which a reliable fatigue life calculation model was established. Subsequently, a fatigue life dataset under various installation conditions of different loads and spatial installation angles was constructed with the model. A DNN was then employed to extract features and construct a fatigue life prediction model. To further optimize the structural life, a multi-parameter dataset was constructed by selecting nearly 40 design dimensions from 7 key components, and a collaborative optimization of the key parameters was conducted based on the DNN prediction model. The results show that the overall fatigue life of the drive shaft assembly is significantly improved by 22.6% after the multi-parameter collaborative optimization. The proposed data-driven method can rapidly and accurately predict, as well as effectively optimize, the fatigue life of drive shafts.

Key words: commercial vehicle driveshaft, fatigue life prediction, data-driven, parameter optimization

CLC Number:

U463.216+2

LIU Xiang, YIN Yuming, WU Zhiwen, DING Xiaoyu, XU Huafu, GAO Jiaqing, YANG Dasheng. Data-driven prediction and multi-parameter optimization of fatigue life of commercial vehicle drive shafts[J]. Journal of Automotive Safety and Energy, 2026, 17(2): 200-208.

Figures/Tables 15

References 16

[1]	江书真, 张博, 舒玉华. 基于深度神经网络的商用车传动轴频率预测[J]. 汽车实用技术, 2019(24): 70-71, 106.
	JIANG Shuzhen, ZHANG Bo, SHU Yuhua. Frequency prediction of commercial vehicle drive shaft based on BP neural network[J]. *Autom Pract Tech*, 2019(24): 70-71, 106. (in Chinese)
[2]	许志鹏, 刘婵, 冯红翠. 基于有限元和RBF神经网络的液压支架前连杆疲劳寿命预测[J]. 机械设计, 2024, 41(1): 110-116.
	XU Zhipeng, LIU Chan, FENG Hongcui. Prediction of fatigue life of the front link of hydraulic suppor-t based on finite element and RBF neural network[J]. *J Mech Desi*, 2024, 41(1): 110-116. (in Chinese)
[3]	张秀华, 刘怀举, 朱才朝, 等. 基于数据驱动的零部件疲劳寿命预测研究现状与发展趋势[J]. 机械传动, 2021, 45(10): 1-14. doi: 10.16578/j.issn.1004.2539.2021.10.001
	ZHANG Xiuhua, LIU Huaiju, ZHU Caichao, et al. Research status and development trend of data-driven fatigue life prediction of parts[J]. *Mech Transm*, 2021, 45(10): 114. (in Chinese)
[4]	刘星荣. 汽车传动系零件寿命及可靠性计算方法研究[J]. 江苏工学院学报, 1993(1): 26-31.
	LIU Xingrong. Research on the calculation method of life and reliability of automobile transmission system parts[J]. *J Jiangsu Instit Tech*, 1993(1): 26-31. (in Chinese)
[5]	刘喆, 陶凤和, 贾长治. 履带车辆传动轴疲劳寿命预测与结构改进[J]. 工程设计学报, 2015, 22(5): 431-437.
	LIU Zhe, TAO Fenghe, JIA Changzhi. Fatigue life prediction and structural improvement of crawler vehicle drive shaft[J]. *J Engi Desig*, 2015, 22(5): 431-437. (in Chinese)
[6]	贾长治, 杜秀菊, 董兆伟, 等. 基于行驶仿真试验的轮边减速器疲劳寿命预测研究[J]. 机械强度, 2014, 36(3): 449-454.
	JIA Changzhi, DU Xiuju, DONG Zhaowei, et al. Research on fatigue life prediction of wheel edge reducer based on driving simulation test[J]. *Mech Stret*, 2014, 36(3): 449-454. (in Chinese)
[7]	周杰, 贾云献, 刘鑫, 等. 基于疲劳寿命的履带车辆侧减速器传动轴结构优化[J]. 机械设计, 2019, 36(4): 82-86.
	ZHOU Jie, JIA Yunxian, LIU Xin, et al. Optimization of drive shaft structure of tracked vehicle side reducer based on fatigue life[J]. *Mech Desig*, 2019, 36(4): 82-86. (in Chinese)
[8]	曹孟杰. 基于机器学习的304不锈钢低周疲劳寿命预测研究[D]. 兰州: 兰州理工大学, 2023.
	CAO Mengjie. Research on prediction of low-cycle fatigue life of 304 stainless steel based on machine learning[D]. Lanzhou: Lanzhou University of Technology, 2023. (in Chinese)
[9]	闫楚良, 郝云霄, 刘克格. 基于遗传算法优化的深度神经网络的材料疲劳寿命预测[J]. 吉林大学学报(工学版), 2014, 44(6): 1710-1715.
	YAN Chuliang, HAO Yunxiao, LIU Kege. Prediction of material fatigue life based on BP neural network optimized by genetic algorithm[J]. J Jilin Univ (Engi Tech Edit), 2014, 44(6): 1710-1715. (in Chinese)
[10]	郑战光, 张剑, 孙腾, 等. 基于多轴载荷相位差的神经网络预测钛合金疲劳寿命[J]. 中国有色金属学报, 2023, 33(3): 781-791.
	ZHENG Zhanguang, ZHANG Jian, SUN Teng, et al. Prediction of fatigue life of titanium alloy by neural network based on multiaxial load phase difference[J]. *J Nonf Meta China*, 2023, 33(3): 781-791. (in Chinese)
[11]	申建国, 汪舟, 卢伟, 等. 有限元模拟和神经网络相结合的喷丸处理SAE9254钢疲劳寿命预测[J]. 机械工程材料, 2024, 48(7): 77-84.
	SHEN Jianguo, WANG Zhou, LU Wei, et al. Prediction of fatigue life of steel SAE9254 shot peening treatment combined with finite element simulation and neural network[J]. *Mech Engi Mater*, 2024, 48(7): 77-84. (in Chinese)
[12]	Vassilopoulos A P, Georgopoulos E F. 5- Novel computational methods for fatigue life modeling of composite materials[M]// Vassilopoulos A P. Fatigue Life Prediction of Composites and Composite Structures. Oxford: Woodhead Publishing, 2010: 139-173.
[13]	魏艳娇, 向清怡. 基于机器学习的焊接接头疲劳寿命预测方法综述[J]. 兵器材料科学与工程, 2025, 48(3): 183-190.
	WEI Yanjiao, XIANG Qingyi. A review of fatigue life prediction methods for welded joints based on machine learning[J]. *Weap Mater Sci Engi*, 2025, 48(3): 183-190. (in Chinese)
[14]	宋宇. 多轴随机载荷下疲劳寿命预测方法的研究[D]. 沈阳: 东北大学, 2012.
	SONG Yu. Research on fatigue life prediction method under multi-axis random load[D]. Shenyang: Northeastern University, 2012. (in Chinese)
[15]	张娟, 戴学宇. 遗传神经网络下超声结节诊断设备异常状态识别研究[J]. 电子设计工程, 2025, 33(11): 154-159.
	ZHANG Juan, DAI Xueyu. Research on abnormal state recognition of ultrasound nodule diagnostic equipment based on genetic neural networks[J]. *Elect Desig Engi*, 2025, 33(11): 154-159. (in Chinese)
[16]	卫昆钰, 李博文, 李晨迪, 等. 基于天牛须搜索算法和内点法的运输类飞机5×5谱编制方法[J]. 航空科学技术, 2024, 35(6): 104-113.
	WEI Kunyu, LI Bowen, LI Chendi, et al. Method for 5×5 spectrum compilation of transport aircraft based on beetle antennae search algorithm and interior point method[J]. *Avia Sci Tech*, 2024, 35(6): 104-113. (in Chinese)

应变片序号	最大扭矩时平均应变/ (μm·m^-1)
应变片序号	0°夹角	5°夹角
NO-01-A	352.12	323.27
NO-01-B	-338.91	-284.93
NO-01-C	-255.90	-204.61
NO-02	-302.77	-270.11
NO-03	-337.78	-301.87

应变片序号	最大扭矩时平均应变/ (μm·m^-1)
应变片序号	0°夹角	5°夹角
NO-01-A	352.12	323.27
NO-01-B	-338.91	-284.93
NO-01-C	-255.90	-204.61
NO-02	-302.77	-270.11
NO-03	-337.78	-301.87

应变位置序号	应变/ (μm·m^-1)
应变位置序号	0°夹角	5°夹角
NO-01	597.62	520.07
NO-02	300.81	265.20
NO-03	341.64	308.20

应变位置序号	应变/ (μm·m^-1)
应变位置序号	0°夹角	5°夹角
NO-01	597.62	520.07
NO-02	300.81	265.20
NO-03	341.64	308.20

零件	尺寸变化范围	仿真样本量	隐藏层	RMSE				相关系数R
零件	尺寸变化范围	仿真样本量	隐藏层	训练	验证	测试	总数据	训练	验证	测试	总数据
输入法兰叉	(±20)%	690	40-25-20-15-6	0.200 7	0.243 7	0.254 2	0.216 3	0.856 4	0.731 5	0.730 3	0.819 3
输出法兰叉		688	45-31-20-12-7	0.090 2	0.152 0	0.137 2	0.109 5	0.901 4	0.602 9	0.674 8	0.832 6
焊接叉		834	50-34-25-15-8	0.238 4	0.353 2	0.334 0	0.284 9	0.880 1	0.663 5	0.756 3	0.814 0
花键叉轴		782	60-32-16-8-4	0.064 0	0.182 3	0.231 1	0.125 8	0.983 5	0.847 8	0.721 0	0.930 2
花键套		642	40-24-20-15-6	0.128 0	0.189 3	0.238 6	0.159 3	0.914 8	0.877 9	0.439 2	0.864 3
十字轴		492	40-30-25-15-6	0.114 4	0.160 1	0.201 2	0.138 1	0.953 7	0.918 3	0.900 4	0.936 9
轴管		120	40-24-15-6	0.340 0	0.357 0	0.148 9	0.321 4	0.721 4	0.824 4	0.909 6	0.745 1