A fatigue assessment for railway maintenance workers based on 3D work posture

doi:10.16265/j.cnki.issn1003-3033.2026.08.0135

Abstract

Abstract:

In order to investigate the relationship between work postures and worker fatigue, a fatigue assessment method based on 3D work posture features is proposed. The study examines the association mechanism among posture characteristics, energy expenditure, and fatigue accumulation during working processes. A digital assessment system for working postures is established using 3D human body keypoints to extract movement features. On this basis, by integrating energy expenditure calculation methods with heat transfer theory, a transformation model from energy expenditure to heat dissipation and subsequently to fatigue level is developed, enabling quantitative characterization of worker fatigue. Based on experimental data from a railway maintenance depot, comparative experiments on fatigue assessment methods and correlation analysis between energy expenditure calculations and physiological indicators are conducted. The results show that the proposed method achieves an average accuracy of 90.22%, outperforming other methods, with precision and F₁ score also significantly higher. Moreover, the assessment results exhibit a significant correlation with physiological indicators. The findings indicate that a quantifiable relationship exists among posture features, energy expenditure, and fatigue levels. The proposed method enables dynamic characterization of fatigue states during working processes, providing a quantitative basis for fatigue identification and process analysis in railway maintenance operations.

Key words: 3D work posture, workers, fatigue levels, railway maintenance, heat dissipation, digital assessment system

CLC Number:

X912安全生理学

Xia Yuxiang, Song Yubo, Mo Junwen. A fatigue assessment for railway maintenance workers based on 3D work posture[J]. China Safety Science Journal, 2026, 36(8): 294-302.

Figures/Tables 6

Fig.1

Fig.2

Fig.3

Table 1

Comparison of the proposed fatigue assessment model with multiple fatigue assessment methods (Experiment 1)

疲劳程度评估		疲劳评估模型				HRV方法^[27]				基于SVM的多参数融合疲劳评估方法^[28]				基于CNN的多参数融合疲劳评估方法^[29]
疲劳程度评估		Acc/ %	Prc/ %	Rec/ %	F₁/ %	Acc/ %	Prc/ %	Rec/ %	F₁/ %	Acc/ %	Prc/ %	Rec/ %	F₁/ %	Acc/ %	Prc/ %	Rec/ %	F₁/ %
组1	P₁	90.28	90.32	87.5	88.89	86.11	85.19	79.31	82.14	84.72	85.29	82.86	84.06	90.28	84.38	93.1	88.52
	P₂	83.59	89.66	89.66	89.66	76.9	80.77	77.78	79.25	86.11	85.19	79.31	82.14	88.89	84.38	90.0	87.1
	P₃	93.06	93.33	90.32	91.8	86.11	80.0	80.0	80.0	84.72	83.33	80.65	81.97	88.89	82.35	93.33	87.5
	P₄	90.28	89.29	86.21	87.72	86.11	85.71	80.0	82.76	88.89	86.96	80.0	83.33	78.48	82.86	90.62	86.57
	P₅	93.06	88.0	91.67	89.8	84.72	80.0	82.76	81.36	87.5	84.62	81.48	83.02	90.28	83.87	92.86	88.14
	P₆	93.06	92.86	89.66	91.23	84.72	83.87	81.25	82.54	84.72	85.29	82.86	84.06	88.89	81.25	92.86	86.67
组2	P₁	93.06	91.3	87.5	89.36	84.72	85.71	77.42	81.36	88.89	87.5	80.77	84.0	90.28	84.38	93.1	88.52
	P₂	94.44	91.3	91.3	91.3	78.4	83.87	81.25	82.54	86.11	84.62	78.57	81.48	90.28	85.29	93.55	89.23
	P₃	80.66	92.31	88.89	90.57	83.33	80.65	80.65	80.65	84.72	87.88	80.56	84.06	91.67	84.62	91.67	88.0
	P₄	91.67	92.59	86.21	89.29	83.33	82.35	82.35	82.35	86.11	82.14	82.14	82.14	90.22	85.19	88.46	86.79
	P₅	94.44	90.91	90.91	90.91	75.78	82.76	82.76	82.76	86.11	85.71	80.0	82.76	91.67	85.71	92.31	88.89
	P₆	91.67	93.33	87.5	90.32	86.11	85.71	80.0	82.76	88.89	86.96	80.0	83.33	88.89	81.25	92.86	86.67
组3	P₁	94.44	90.91	90.91	90.91	75.19	80.65	80.65	80.65	77.87	84.62	78.57	81.48	75.27	82.86	90.62	86.57
	P₂	80.24	93.33	90.32	91.8	86.01	85.19	79.31	82.14	75.08	83.87	81.25	82.54	91.67	85.19	92.0	88.46
	P₃	91.67	93.33	87.5	90.32	86.18	84.62	78.57	81.48	79.92	86.67	81.25	83.87	76.1	83.33	90.91	86.96
	P₄	91.67	93.26	87.41	90.32	79.38	80.0	80.0	80.0	88.89	87.5	80.77	84.0	88.93	84.85	90.32	87.5
	P₅	93.15	92.0	88.46	90.2	83.33	84.85	80.01	82.35	88.82	87.5	80.77	84.0	88.92	84.85	90.32	87.5
	P₆	90.28	90.62	87.88	89.23	86.15	82.76	82.76	82.76	86.15	87.1	81.82	84.38	90.22	83.87	92.86	88.14
组4	P₁	90.28	90.32	87.46	88.89	87.5	83.33	80.03	81.63	79.46	84.0	80.77	82.35	91.67	83.33	90.91	86.96
	P₂	93.11	88.0	91.67	89.8	77.26	81.25	81.25	81.25	87.5	87.5	77.78	82.35	90.18	84.85	93.33	88.89
	P₃	81.09	90.91	90.91	90.91	77.24	84.62	81.48	83.02	87.5	84.62	81.48	83.02	90.37	84.85	93.33	88.89
	P₄	94.44	91.3	91.3	91.3	76.59	83.33	80.65	81.97	84.79	87.1	79.41	83.08	90.3	86.67	89.66	88.14
	P₅	91.76	93.1	87.1	90.0	87.5	83.33	80.07	81.63	86.06	86.67	81.25	83.87	91.68	85.19	92.0	88.46
	P₆	87.01	92.59	89.29	90.91	79.83	83.33	80.65	81.97	86.09	86.21	80.65	83.33	90.24	85.29	93.55	89.23
组5	P₁	93.15	92.59	89.29	90.88	77.8	81.48	78.57	80.0	86.13	84.69	78.57	81.48	88.87	84.88	90.32	87.5
	P₂	79.9	90.0	87.1	88.52	86.18	84.0	77.78	80.77	86.09	87.88	82.86	85.29	88.82	82.76	88.89	85.71
	P₃	90.29	89.66	86.67	88.14	87.5	83.38	80.05	81.63	84.74	87.44	80.0	83.58	90.19	86.67	89.66	88.14
	P₄	90.29	89.66	86.67	88.14	83.33	82.86	82.86	82.86	87.5	83.33	79.95	81.63	88.96	84.79	90.32	87.6
	P₅	93.01	92.31	88.89	90.57	84.72	81.48	78.57	80.0	84.77	85.29	82.86	84.06	91.77	86.21	92.59	89.29
	P₆	91.68	88.89	88.89	88.89	76.55	84.62	81.48	83.02	84.73	87.1	79.41	83.08	90.33	84.86	93.33	88.89
平均		90.22	91.27	88.84	90.02	82.49	83.06	80.34	81.65	85.32	85.82	80.62	83.12	88.77	84.36	91.65	87.85

Table 1

Fig.4

Table 2

References 30

[1]	谢泽, 张可新, 李金波. 高速铁路运营安全保障技防体系研究[J]. 铁道运输与经济, 2024, 46(7): 151-158.
	Xie Ze, Zhang Kexin, Li Jinbo. Technical defense system for operational safety guarantee of high speed railway[J]. Railway Transport and Economy, 2024, 46(7): 151-158.
[2]	王健, 邢科家, 王玺. 预防性工间休策略应对铁路运输岗位疲劳作业的研究[J]. 铁道科学与工程学报, 2024, 21(8): 3119-3130.
	Wang Jian, Xing Kejia, Wang Xi. Study on preventive break strategies to cope with fatigue operation of railway transport post[J]. Journal of Railway Science and Engineering, 2024, 21(8): 3119-3130.
[3]	Santos L F F M, Melicio R. Stress, pressure and fatigue on aircraft maintenance personal[J]. International Review of Aerospace Engineering (IREASE), 2019, 12(1): 35-45. doi: 10.15866/irease.v12i1.14860
[4]	陶克雪, 陈建武, 蒋引, 等. 作业人员疲劳测量量表应用研究分析[J]. 中国安全生产科学技术, 2023, 19(7): 186-191.
	Tao Kexue, Chen Jianwu, Jiang Yin, et al. Application research analysis of fatigue measurement scale for workers[J]. Journal of Safety Science and Technology, 2023, 19(7): 186-191.
[5]	王健, 高扬, 邢科家, 等. 铁路运输岗位疲劳研究综述[J]. 中国铁路, 2023(3): 16-25.
	Wang Jian, Gao Yang, Xing Kejia, et al. Overview of research on fatigue of railway transport posts[J]. China Railway, 2023(3): 16-25.
[6]	陈建武, 毕春波, 廖海江, 等. 作业疲劳测量方法对比研究[J]. 中国安全生产科学技术, 2011, 7(5): 63-66.
	Chen Jianwu, Bi Chunbo, Liao Haijiang, et al. Comparative research on measurement methods of work fatigue[J]. Journal of Safety Science and Technology, 2011, 7(5): 63-66.
[7]	Xu Shuling, Hall N G. Fatigue, personnel scheduling and operations: review and research opportunities[J]. European Journal of Operational Research, 2021, 295(3): 807-822. doi: 10.1016/j.ejor.2021.03.036
[8]	Bustos D, Guedes J C, Vaz M P, et al. Non-invasive physiological monitoring for physical exertion and fatigue assessment in military personnel: a systematic review[J]. International Journal of Environmental Research and Public Health, 2021, 18(16): DOI:10.3390/ijerph18168815.
[9]	陆荣秀, 张笔豪, 莫振龙. 基于脸部特征和头部姿态的疲劳检测方法[J]. 系统仿真学报, 2022, 34(10): 2279-2292. doi: 10.16182/j.issn1004731x.joss.21-0583
	Lu Rongxiu, Zhang Bihao, Mo Zhenlong. Fatigue detection method based on facial features and head posture[J]. Journal of System Simulation, 2022, 34(10): 2279-2292. doi: 10.16182/j.issn1004731x.joss.21-0583
[10]	Putri A A, Diaturrochma M, Putri A S, et al. Relationship between nutritional status and work posture with work fatigue on crane operators[J]. The Indonesian Journal of Occupational Safety and Health, 2022, 11(2): 266-274. doi: 10.20473/ijosh.v11i2.2022.266-274
[11]	白仲航, 项钲, 谭昭芸, 等. 自适应神经模糊推理系统优化的快速上肢评估方法[J]. 计算机集成制造系统, 2024, 30(5): 1643-1656.
	Bai Zhonghang, Xiang Zheng, Tan Zhaoyun, et al. Rapid upper limb assessment method based on adaptive neuro fuzzy inference system optimization[J]. Computer Integrated Manufacturing Systems, 2024, 30(5): 1643-1656.
[12]	Wang Luoluo, Zhang Hao, Chen Junliang, et al. Analysis and exploration of automotive fatigue driving detection technology based on multimode fusion[J]. Academic Journal of Engineering and Technology Science, 2023, 6(7): 24-29.
[13]	刘国丹, 张瑶, 胡鹏城, 等. 基于红外热成像技术的着装热阻计算与热舒适:夏季室内不同活动水平下的人员研究[J]. 建筑科学, 2024, 40(8): 139-149.
	Liu Guodan, Zhang Yao, Hu Pengcheng, et al. Thermal resistance calculation and thermal comfort of dressing based on infrared thermography: a study of people at different activity levels indoors in summer[J]. Building Science, 2024, 40(8): 139-149.
[14]	胡钦太, 伍文燕, 冯广, 等. 深度学习支持下多模态学习行为可解释性分析研究[J]. 电化教育研究, 2021, 42(11): 77-83.
	Hu Qintai, Wu Wenyan, Feng Guang, et al. A study on interpretable analysis of multimodal learning behavior supported by deep learning[J]. E-Education Research, 2021, 42(11): 77-83.
[15]	Bayoudh K. A survey of multimodal hybrid deep learning for computer vision: architectures, applications, trends, and challenges[J]. Information Fusion, 2024, 105: DOI:10.1016/j.inffus.2023.102217.
[16]	Chen Siqi. SCMotion: predicting human motion through spatial dependence and center anchoring[C]. 2024 7^th International Conference on Computer Information Science and Application Technology (CISAT), 2024: 1226-1230.
[17]	Vallez L J, Plourde B D, Abraham J P. A new computational thermal model of the whole human body: applications to patient warming blankets[J]. Numerical Heat Transfer, Part A: Applications, 2016, 69(3): 227-241. doi: 10.1080/10407782.2015.1080573
[18]	Nakata T, Mori T, Sato T. Analysis of impression of robot bodily expression[J]. Journal of Robotics and Mechatronics, 2002, 14(1): 27-36. doi: 10.20965/jrm.2002.p0027
[19]	Tanaka M, Tajima S, Mizuno K, et al. Frontier studies on fatigue, autonomic nerve dysfunction, and sleep-rhythm disorder[J]. The Journal of Physiological Sciences, 2015, 65(6): 483-498. doi: 10.1007/s12576-015-0399-y
[20]	Li Wang. Evaluation of running intensity and fatigue degree based on human physiological information[J]. Mobile Information Systems, 2022, 2022(1): DOI:10.1155/2022/6488312.
[21]	Rahmatika N N, Sutawardana J H, Hakam M. Hubungan manajemen energi dengan kelelahan pada pasien diabetes melitus tipe 2[J]. Jurnal Keperawatan Priority, 2022, 5(1): 118-123. doi: 10.34012/jukep.v5i1
[22]	Ahmed S, Babski-Reeves K, DuBien J, et al. A proposed relationship between time and load to quantify fatigue[C]. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2014, 58(1): 1556-1560.
[23]	Matthews G, Desmond P A, Hancock P A, et al. The handbook of operator fatigue[M]. Burlington, VT: Ashgate Pub. Company, 2012: 123-170.
[24]	Micklewright D, St Clair Gibson A, Gladwell V, et al. Development and validity of the rating-of-fatigue scale[J]. Sports Medicine, 2017, 47(11): 2375-2393. doi: 10.1007/s40279-017-0711-5 pmid: 28283993
[25]	徐清浩, 陈浩, 陈琳, 等. 中等运动强度下人体热感觉与生理响应及其相关性实验研究[J]. 建筑科学, 2021, 37(8): 67-78, 86.
	Xu Qinghao, Chen Hao, Chen Lin, et al. Experimental study on the correlation between thermal sensation and physiological response at moderate exercise intensity[J]. Building Science, 2021, 37(8): 67-78, 86.
[26]	Pind R, Hofmann P, Mäestu E, et al. Increases in RPE rating predict fatigue accumulation without changes in heart rate zone distribution after 4-week low-intensity high-volume training period in high-level rowers[J]. Frontiers in Physiology, 2021, 12: DOI:10.3389/fphys.2021.735565.
[27]	Ni Zhiqiang, Sun Fangmin, Li Ye. Heart rate variability-based subjective physical fatigue assessment[J]. Sensors (Basel, Switzerland), 2022, 22(9): DOI:10.3390/s22093199.
[28]	胡峰松, 程哲坤, 徐青云, 等. 基于多特征融合的疲劳驾驶状态识别方法研究[J]. 湖南大学学报(自然科学版), 2022, 49(4): 100-107.
	Hu Fengsong, Cheng Zhekun, Xu Qingyun, et al. Research on fatigue driving state recognition method based on multi-feature fusion[J]. Journal of Hunan University (Natural Sciences), 2022, 49(4): 100-107.
[29]	Zhao Shuaijie, Du Aimin, Han Yeyang, et al. Multimodal features fusion for driver fatigue detection based on CNN-GTN learning[C]. 2023 7^th CAA International Conference on Vehicular Control and Intelligence (CVCI), 2023: 1-6.
[30]	Ni Weiguang, Zhang Yuxin, Li Xinyi, et al. A study on the relationship between RPE and sEMG in dynamic contraction based on the GPR method[J]. Electronics, 2022, 11(5): DOI: 10.3390/electronics11050691.

试验相关指标	作业强度			时间段
分类	高/2.8~3.6 Mets	中/2.2~2.7 Mets	低/1.3~2.1 Mets	前半夜	后半夜
RPE	0.709	0.785	0.745	0.785	0.710
HR	0.795	0.841	0.807	0.841	0.853
sEMG	0.862	0.852	0.813	0.852	0.848
VO₂	0.833	0.812	0.823	0.812	0.811
VCO₂	0.857	0.795	0.834	0.795	0.818
MV	0.808	0.834	0.839	0.834	0.841