基于近似马尔可夫毯与GBDT的煤矿职业健康噪声风险分类

doi:10.16265/j.cnki.issn1003-3033.2025.09.1075

摘要/Abstract

摘要：

为精准判断综采工作面噪声对作业人员的健康损害程度,采用人-机-环-管系统理论结合Fisher Score、最大信息系数及近似马尔可夫毯方法,确定噪声职业健康的关键影响因素;构建基于梯度提升决策树(GBDT)算法的煤矿噪声风险分类预测模型,并以Kappa系数及其准确率作为模型效率的指标,来对比验证本算法模型的准确性。结果表明:综采工作面噪声职业健康损害情况与个体状况、设备配置、环境因素及职业健康管理等因素密切相关,其中,岗位类别、个体年龄、工龄时长、防护意识、设备自动化程度、噪声监测点合格率、噪声暴露量、混响时间和管理机构及人员为职业健康风险分类预测的关键指标;基于GBDT构建的煤矿噪声职业健康风险分类预测模型准确率最大达99.6%,平均准确率和Kappa系数分别为98.3%和0.958;计算确定6种综采工作面噪声职业健康风险分类预测模型评估准确率次序为:GBDT > 遗传算法优化随机森林算法(GA-RF)> 粒子群算法优化最小二乘支持向量机算法(PSO-LSSVM)> 随机森林算法(RF)> 支持向量机算法(SVM)> 决策树。

关键词: 近似马尔可夫毯, 梯度提升决策树(GBDT), 煤矿噪声, 噪声健康风险分类, 职业健康, Fisher Score

Abstract:

To accurately assess the impact of noise on the health of workers in fully mechanized mining face, the key influencing factors of noise occupational health were determined by using the theory of man-machine-environment-management system, combined with Fisher Score, maximum information coefficient and approximate Markov blanket method. The prediction model of coal mine noise risk classification based on the GBDT algorithm was constructed, and the Kappa coefficient and its accuracy were used as the index of model efficiency to compare and verify the accuracy of the model. The results show that the occupational health damage of noise in fully mechanized mining face is closely related to individual status, equipment configuration, environmental factors and occupational health management. Among these, job category, individual age, length of service, protection awareness, degree of equipment automation, pass rate of noise monitoring points, noise exposure, reverberation time and management institutions and personnel are the key indicators of occupational health risk classification and prediction. The accuracy of the occupation health risk classification prediction model of coal mine noise based on GBDT is up to 99.6%, and the average accuracy and Kappa coefficient are 98.3% and 0.958, respectively. The evaluation accuracy of six prediction models for noise occupational health risk classification in fully mechanized mining face is ranked as follows: GBDT > Genetic Algorithm optimization Random Forest(GA-RF)> Particle Swarm optimization Least Squares Support Vector Machine(PSO-LSSVM)> Random Forest(RF)> Support Vector Machine(SVM)> Decision tree.

Key words: approximate Markov blanket, gradient boosting decision tree (GBDT), coal mine noise, noise health risk classification, occupational health, Fisher Score

中图分类号:

X966

高晓旭, 田佳可, 高璐, 杜芦, 范萌杰. 基于近似马尔可夫毯与GBDT的煤矿职业健康噪声风险分类[J]. 中国安全科学学报, 2025, 35(9): 253-262.

GAO Xiaoxu, TIAN Jiake, GAO Lu, DU Lu, FAN Mengjie. Noise risk classification of coal mine occupational health based on approximate Markov blanket and GBDT[J]. China Safety Science Journal, 2025, 35(9): 253-262.

图/表 15

图1

表1

图2

表2

表3

图3

图4

图5

表4

表5

图6

图7

图8

图9

图10

参考文献 17

[1]	顾大钊, 李全生. 基于井下生态保护的煤矿职业健康防护理论与技术体系[J]. 煤炭学报, 2021, 46(3): 950-958.
	GU Dazhao, LI Quansheng. Theoretical framework and key technologies of underground ecological protection based on coal mine occupational health prevention[J]. Journal of China Coal Society, 2021, 46(3): 950-958.
[2]	SUN Kan, AZMAN A S. Evaluating hearing loss risks in the mining industry through MSHA citations[J]. Journal of Occupational and Environmental Hygiene, 2018, 15(3): 246-262. doi: 10.1080/15459624.2017.1412584 pmid: 29200378
[3]	SUN Kan, AZMAN A S, CAMARGO H E, et al. Risk assessment of recordable occupational hearing loss in the mining industry[J]. International Journal of Audiology, 2019, 58(11): 761-768. doi: 10.1080/14992027.2019.1622041 pmid: 31282793
[4]	常波峰, 黄伟, 刘宽, 等. 陕西省某煤矿企业噪声职业健康风险评估[J]. 公共卫生与预防医学, 2024, 35(1): 70-73.
	CHANG Bofeng, HUANG Wei, LIU Kuan, et al. Occupational health risk assessment of noise in a coal mining enterprise in Shaanxi province[J]. Journal of Public Health and Preventive Medicine, 2024, 35(1): 70-73.
[5]	王海椒, 邹晓雪, 刘丽华, 等. 模糊层次分析法在煤矿生产性噪声危害分级评价中的应用[J]. 中国工业医学杂志, 2019, 32(4): 286-289.
	WANG Haijiao, ZOU Xiaoxue, LIU Lihua, et al. Application of fuzzy analytic hierarchy process in classification and evaluation of productive noise hazard[J]. Chinese Journal of Industrial Medicine, 2019, 32(4): 286-289.
[6]	MILAD A, SAEID Y, ALI A. Developing and validating a risk assessment method for noise-induced hearing loss in workers[J]. Heliyon, 2024, 10(22):DOI: 10.1016/J.HELIYON.2024.E40475.
[7]	ASGHARI M, GORJI R, MORADZADEH R, et al. A risk model for occupational noise-induced hearing loss in workers[J]. Work, 2023, 77(3): 1017-1022. doi: 10.3233/WOR-230181 pmid: 37781851
[8]	QI Haihui, DU Yiyi, TIAN Yu, et al. Research on early warning model of the hearing loss of workers exposed to noise[J]. Chinese Journal of Industrial Hygiene and Occupational Diseases, 2023, 41(1): 47-51.
[9]	王坤. 机械加工车间环境质量及其职业健康损害评价研究[D]. 重庆: 重庆大学, 2019.
	WANG Kun. Study on environmental quality and occupational health damage assessment in mechanical processing workshop[D]. Chongqing: Chongqing University, 2019.
[10]	陈春艳, 龚志起. 钢构件生产车间噪声健康损害评价[J]. 科学技术创新, 2021(10): 32-35.
[11]	李丹丹. 不同类型巷道中噪声危害评价研究[D]. 廊坊: 华北科技学院, 2020.
	LI Dandan. Study on noise hazard evaluation in different types of laneways[D]. Langfang: North China Institute of Science and Technology, 2020.
[12]	赵玲, 龚加兴, 黄大荣, 等. 基于Fisher Score与最大信息系数的齿轮箱故障特征选择方法[J]. 控制与决策, 2021, 36(9): 2234-2240.
	ZHAO Ling, GONG Jiaxing, HUANG Darong, et al. Fault feature selection method of gearbox based on Fisher Score and maximum information coefficient[J]. Control and Decision, 2021, 36(9): 2 234-2 240.
[13]	贾改妮, 蒋方, 杨明, 等. 煤矿井下不同作业场所的职业健康损害研究[J]. 中国安全科学学报, 2023, 33(4): 221-229. doi: 10.16265/j.cnki.issn1003-3033.2023.04.1759
	JIA Gaini, JIANG Fang, YANG Ming, et al. Research on occupational health damage in different working places in underground coal mine[J]. China Safety Science Journal, 2023, 33(4): 221-229. doi: 10.16265/j.cnki.issn1003-3033.2023.04.1759
[14]	RESHEF D N, RESHEF Y A, FINUCANE H K, et al. Detecting novel associations in large data sets[J]. Science, 2011, 334(6062): 1518-1524. doi: 10.1126/science.1205438 pmid: 22174245
[15]	马陇飞, 萧汉敏, 陶敬伟, 等. 基于梯度提升决策树算法的岩性智能分类方法[J]. 油气地质与采收率, 2022, 29(1): 21-29.
	MA Longfei, XIAO Hanmin, TAO Jingwei, et al. Intelligent lithology classification method based on GBDT algorithm[J]. Petroleum Geology and Recovery Efficiency, 2022, 29(1): 21-29.
[16]	高晓旭, 高璐, 潘相旭, 等. 基于t-SSA-BP的煤矿噪声职业健康损害预测[J]. 中国安全科学学报, 2023, 33(12): 214-222. doi: 10.16265/j.cnki.issn1003-3033.2023.12.2068
	GAO Xiaoxu, GAO Lu, PAN Xiangxu, et al. Prediction of occupational health damage caused by coal mine noise based on t-SSA-BP[J]. China Safety Science Journal, 2023, 33(12): 214-222. doi: 10.16265/j.cnki.issn1003-3033.2023.12.2068
[17]	韩启迪, 张小桐, 申维. 基于梯度提升决策树(GBDT)算法的岩性识别技术[J]. 矿物岩石地球化学通报, 2018, 37(6): 1173-1180.
	HAN Qidi, ZHANG Xiaotong, SHEN Wei. Lithology identification technology based on gradient boosting decision tree (GBDT) algorithm[J]. Bulletin of Mineralogy, Petrology and Geochemistry, 2018, 37(6): 1 173-1 180.

指标类型		目标类型	等级界限(可能遭受的噪声职业健康风险)
一级	二级	目标类型	Ⅰ(严重)	Ⅱ(可接受)	Ⅲ(潜在)
个体状况	个体年龄X₂/岁	效益型	[20,30)	[30,50]	(50,60]
	工龄时长X₃/a	成本型	(20,40]	[2,20]	(0,2)
	文化水平X₄/分	效益型	1	[2,4]	5
	专业技能X₅/分	效益型	1	[2,4]	5
	危害认知X₆/%	效益型	[0,80)	[80,95)	[95,100]
	防护意识X₇/%	效益型	[0,80)	[80,95)	[95,100]
设备配置	设备自动化程度X₈/%	效益型	[0,80)	[80,95)	[95,100]
	设备布局X₉/%	效益型	[0,80)	[80,95)	[95,100]
	工艺流程X₁₀/%	效益型	[0,80)	[80,95)	[95,100]
	噪声降吸装置X₁₁/%	效益型	[0,80)	[80,95)	[95,100]
	消声设备使用率X₁₂/%	效益型	[0,80)	[80,95)	[95,100]
	个体防护用具X₁₃/%	效益型	[0,80)	[80,95)	[95,100]
环境因素	距噪声源距离X₁₄/m	效益型	[0,20)	[20,90]	(90,200]
	噪声监测点合格率X₁₅/%	效益型	[0,80)	[80,95)	[95,100]
	噪声暴露量X₁₆/dB	成本型	[100,200]	[85,100)	[0,85)
职业健康管理	管理机构及人员X₁₈/%	效益型	[0,80)	[80,95)	[95,100]
	职业危害监测X₁₉/%	效益型	[0,80)	[80,95)	[95,100]
	职业卫生培训率X₂₀/%	效益型	[0,80)	[80,95)	[95,100]
	警告标识设置X₂₁/%	效益型	[0,80)	[80,95)	[95,100]
	职业病危害告知率X₂₂/%	效益型	[0,80)	[80,95)	[95,100]

因素	Fisher Score	因素	Fisher Score
X₁	0.289	X₁₂	0.322
X₂	0.250	X₁₃	0.662
X₃	0.827	X₁₄	0.176
X₄	0.165	X₁₅	1.592
X₅	0.147	X₁₆	3.663
X₆	0.016	X₁₇	0.228
X₇	0.752	X₁₈	0.640
X₈	0.093	X₁₉	0.064
X₉	0.051	X₂₀	0.014
X₁₀	0.008	X₂₁	0.013
X₁₁	0.306	X₂₂	0.014

因素	重要度	因素	重要度
X₁	0.246	X₁₂	0.052
X₂	0.264	X₁₃	0.186
X₃	0.267	X₁₄	0.131
X₄	0.157	X₁₅	0.379
X₅	0.018	X₁₆	0.706
X₇	0.268	X₁₇	0.206
X₈	0.246	X₁₈	0.356
X₉	0.053	X₁₉	0.109
X₁₁	0.072	—	—

数据编号	工种	监测地点	年龄/ 岁	工龄/ a	防护意识/%	设备自动化程度/%	监测点合格率/%	噪声暴露量 L_EX,8h/dB(A)	混响时间/s	管理机构及人员比例/%
1	采煤机司机	采煤机司机操作位	46	15	82	86	90	94.1	0.52	88
2	超前支护工	超前支护工操作位	49	18	86	85	79	82.1	0.65	81
3	刮板机司机	刮板机司机操作位	50	20	94	81	86	93.5	0.49	75
4	转载机司机	转载机司机操作位	37	11	81	87	93	92.7	0.73	83
5	破碎机司机	破碎机旁	42	13	89	88	85	95.3	0.55	90
6	乳化泵工	设备列车旁	45	10	90	93	89	96.4	0.62	85
7	带式输送机司机	带式输送机司机操作位	35	8	79	89	83	85.8	0.83	91
︙	︙	︙	︙	︙	︙	︙	︙	︙	︙	︙
612	带式输送机司机	带式输送机司机操作位	52	19	92	82	81	88.6	0.77	87

超参数	数值范围
迭代次数	[100,1 000],步长为100
模型学习率	0.001,0.01,0.02,0.03,…,0.1
树最大深度	[1,9],步长为2
划分所需最小样本数	[1,5],步长为1