煤炭产业是我国经济的重要支柱，肩负着确保国家能源安全和经济发展的重要使命。然而，煤矿安全事故的发生给煤炭行业和社会稳定带来了巨大挑战。经调查和统计分析发现，运输事故在总事故中占据较大比例，而胶轮车驾驶员的不安全行为是导致运输事故发生的直接原因。胶轮车驾驶员在不同强度的作业任务和矿井环境下处于不同认知负荷状态，长期高负荷状态会对其心理、生理和行为产生不利影响，促使不安全行为的发生。因此，胶轮车驾驶员在不同认知负荷状态下的行为应当被重视，并有针对性地进行干预，以预防不安全行为，降低煤矿事故发生率。主要研究内容和结论如下：

（1）建立研究模型和假设，设计并完成了基于认知负荷的驾驶员不安全行为实验。通过对当前国内外认知负荷与行为之间的关系已有的研究成果进行总结，建立认知负荷与不安全行为影响的研究模型和假设，设计基于认知负荷的驾驶员不安全行为的研究实验，采集并统计分析了被试在低、中和高三种认知负荷下的生理信号、行为学数据和主观量表得分，利用美国航天局任务负荷指数（NASA-TLX）量表获取被试的主观认知负荷得分，基于NASA-TLX量表得分验证了N-back范式可效诱发被试不同程度的认知负荷。

（2）分析实验数据，验证了研究假设并确定了可用于表征胶轮车驾驶员不安全行为的指标。使用SPSS26.0对各指标进行显著性、平均值和标准差分析，发现随着认知负荷难度的上升，心电指标中的LF/HF、HR指标、NASA量表得分、行为指标（反应时间、正确率）、眼动行为指标（瞳孔直径、注视时长）显著增加，心电指标（SDNN、MEANRR、RMSSD、VLF、HF）和注视次数显著降低。根据数据分析发现在不同认知负荷状态下，每位被试的生理指标和行为指标存在显著差异，从而验证了研究假设。利用Lasso回归模型对特征值进行选取，结果显示只有低频功率指标（LF）对不安全行为的重要性为0，剔除该指标，其余指标可作为胶轮车驾驶员不安全行为的表征指标。

（3）使用K-means聚类分析法对不安全行为进行了分类。对实验所获得的行为学数据、NASA-TLX得分和生理指标进行聚类，根据聚类结果，将胶轮车驾驶员不安全行为等级划分为三类，分别为：安全行为、可能发生不安全行为、不安全行为。

（4）利用机器学习建立了胶轮车驾驶员基于认知负荷的不安全行为预测模型。将筛选出来的不安全行为表征指标作为输入向量，不安全行为类别作为输出向量，基于支持向量机、K近邻和随机森林分别构胶轮车驾驶员不安全行为预测模型。通过计算三种分类器的准确率、精确率、召回率和F1值，综合比较发现随机森林算法整体预测效果最好，准确率达到83.3%，说明该模型能够较好的预测不安全行为。故本研究选择随机森林预测模型对胶轮车驾驶员的不安全行为进行预测，为后续胶轮车驾驶员不安全行为预警和管控提供理论依据。

The coal industry is an important pillar of China's economy and shoulders an important mission to ensure national energy security and economic development. However, the occurrence of coal mine safety accidents has brought great challenges to the coal industry and social stability. Through investigation and statistical analysis, it is found that transportation accidents occupy a large proportion in the total accidents, and the unsafe behavior of the drivers of rubber wheel vehicles is the direct cause of transportation accidents. The drivers of rubber wheel vehicles are in different cognitive load states under different intensity tasks and mine environment, and the long-term high load state will have adverse effects on their psychology, physiology and behavior, and promote the occurrence of unsafe behaviors. Therefore, the behavior of rubber wheel drivers under different cognitive load states should be paid attention to, and targeted intervention should be carried out to prevent unsafe behavior and reduce the incidence of coal mine accidents. The main research contents and conclusions are as follows:

(1) Establish the research model and hypothesis, design and complete the driver unsafe behavior experiment based on cognitive load. Based on a summary of the existing research results on the relationship between cognitive load and behavior at home and abroad, a research model and hypothesis on the impact of cognitive load and unsafe behavior were established, a research experiment on drivers' unsafe behavior based on cognitive load was designed, and physiological signals, behavioral data and subjective scale scores of subjects under low, middle and high cognitive load were collected and statistically analyzed. The NASA Mission Load Index (NASA-TLX) scale was used to obtain subjects' subjective cognitive load scores. Based on the NASA-TLX scale score, it was verified that N-back paradigm could effectively induce subjects' cognitive load to different degrees.

(2) By analyzing the experimental data, the research hypothesis is verified and the indexes which can be used to characterize the unsafe behavior of the drivers of rubber wheels are determined. SPSS26.0 was used to analyze the significance, mean value and standard deviation of each indicator. It was found that LF/HF, HR index, NASA scale score, behavior index (reaction time, accuracy rate) and eye movement behavior index (pupil diameter and fixation length) of ECG index increased significantly with the increase of cognitive load difficulty. Ecg parameters (SDNN, MEANRR, RMSSD, VLF, HF) and fixation times were significantly reduced. According to the data analysis, it was found that there were significant differences in the physiological and behavioral indicators of each subject under different cognitive load states, thus verifying the research hypothesis. Lasso regression model was used to select the characteristic values, and the results showed that only the low frequency power index (LF) was 0 in importance to unsafe behavior. If LF was removed, the other indexes could be used as the characteristic indexes of unsafe behavior of rubber wheel drivers.

(3) Unsafe behaviors are classified by K-means clustering analysis. The behavioral data, NASA-TLX scores and physiological indicators obtained in the experiment were clustered. According to the clustering results, unsafe behavior levels of rubber wheel drivers were divided into three categories: safe behavior, possible unsafe behavior, and unsafe behavior.

(4) A cognitive load based unsafe behavior prediction model for rubber bike drivers was established using machine learning. Unsafe behavior indexes were selected as input and unsafe behavior categories as output. Unsafe behavior prediction models were based on support vector machine, K-nearest neighbor and random forest. By calculating the accuracy rate, accuracy rate, recall rate and F1 value of the three classifiers, it is found that random forest algorithm has the best prediction effect, and the accuracy rate reaches 83.3%, indicating that this model can better predict unsafe behavior. Therefore, this study selects the random forest prediction model to predict the unsafe behavior of rubber bicycle drivers, and provides a theoretical basis for the subsequent warning and control of unsafe behavior of rubber bicycle drivers.

[1] 薛黎明, 王豪杰, 朱兵兵, 等. 煤炭资源可持续力评价与系统协调发展分析[J]. 经济地理, 2020, 40(01): 114-124.

[2] 郑剑娜. 煤炭企业标准化管理对产品生产的影响[J]. 内蒙古煤炭经济, 2020(14): 90-91.

[3] 周坪, 马国庆, 周公博等. 智能化带式输送机健康监测技术研究综述[J]. 仪器仪表学报, 2023, 44(12): 1-21.

[4] 邓永胜. 煤矿井下无轨胶轮车的现状及应用[J]. 矿业装备,2023(02):165-167.

[5] 程磊, 孙洁. 2016—2022年我国煤矿事故统计与规律分析[J]. 煤炭工程, 2023, 55(11): 125-129.

[6] 张景林, 林柏泉. 安全学原理[M]. 北京: 中国劳动社会保障出版社, 2009: 86-90.

[7] 李盼. 事故体验对矿工不安全行为影响研究[D]. 西安: 西安建筑科技大学, 2021.

[8] Heinrich, H. W. Industrial Accident Prevention[M]. New York: Mc Graw-hill, 1979.

[9] 田水承, 李红霞, 王莉. 3类危险源与煤矿事故防治[J]. 煤炭学报, 2006(06): 706-710.

[10] Alaimo B. Unsafe behavior or unsafe condition-That is the question?[J]. Journal of Chemical Health and Safety, 2006, 13(1): 48-49.

[11] 田水承, 丁洋, 匡秘姈. 疲劳中介效应下矿工心理因素对不安全行为的影响[J]. 西安科技大学学报, 2023, 43(01): 47-54.

[12] 王露露, 尘兴邦, 袁嘉淙, 等. 职业心理学视域下矿工不安全行为研究综述[J]. 中国安全科学学报, 2023, 33(01): 48-55.

[13] 李磊, 景兴鹏, 田水承. 基于SEM的矿工不安全行为形成机理研究[J]. 煤矿安全, 2016, 47(07): 234-236.

[14] 沈剑, 李红霞. 矿工作业疲劳对煤矿险兆事件的影响机理——基于情感耗竭中介变量的分析[J]. 安全与环境学报, 2019, 19(02): 527-534.

[15] 任志成, 孔德中, 宋高峰, 等. 基于GRA和AHP的煤矿一般事故防控研究[J]. 矿业研究与开发, 2023, 43(12): 131-137.

[16] 叶贵, 杨丽娟, 汪红霞, 等. 建筑工人不安全行为影响因素的多层递阶结构模型研究[J]. 安全与环境工程, 2019, 26(02): 129-134.

[17] 朱艳娜, 衡连伟, 何刚, 等. 煤矿员工不安全行为影响因素作用效应分析[J]. 矿业安全与环保, 2019, 46(01): 104-108.

[18] Wang D, Wang X, Xia N. How safety-related stress affects workers’ safety behavior:the moderating role of psychological capital[J]. Safety Science, 2018, 103(4): 247-259.

[19] Sadeghi Y M, Rastegarzadeh E, Soltanzadeh A, et al. Modeling the occurrence of unsafe behaviors based on safety climate and organizational leadship style dimensions in oil platforms using the fuzzy logic approach[J]. Concurrency and Computation:Practice and Experience, 2021, 34(4): 128-132.

[20] Jung M ,Lim S, Chi S. Impact of work environment and occupational stress on safety behavior of individual construction workers.[J]. Multidisciplinary Digital Publishing Institute, 2020, 17(22): 83-89.

[21] Zou Patrick X. W,Wang Ruili. Overcoming the Effect of Young Workers’ Rebellious Psychology on Unsafe Behavior in Construction[M]. Springer Nature Singapore, 2023: 1767-1782.

[22] Kumar Deepak,Bhattacharjee Ram Madhab. Reducing workplace unsafe behaviour using risk classification, profiling, risk tolerance approach[J]. Heliyon, 2023, 9(3): e13969-e13969.

[23] 屈婷. 矿工不安全行为量表设计及实证研究[D]. 西安: 西安科技大学, 2013.

[24] 陆宝军, 傅贵, 王劲松, 等. 苏丹项目中资企业员工不安全行为心理测量[J]. 安全与环境学报, 2013, (02): 222-226.

[25] 张佳彩. 煤矿工人不安全行为测量与实践[D]. 山东: 山东科技大学. 2011.

[26] 赵泓超. 基于生理—心理测量的矿工不安全行为实验研究[D]. 西安: 西安科技大学, 2012.

[27] 杨妍. 矿工疲劳与不安全行为的实验研究—基于神经管理学的视角[D]. 西安: 西安科技大学, 2013.

[28] 田水承, 梁清, 王莉, 等. 噪声与矿工行为安全关系研究及防控对策[J]. 西安科技大学学报, 2015, 35(05): 555-560.

[29] Chandler P, Sweller J. Cognitive Load Theory and the Format of Instruction[J]. Cognition & Instruction, 1991, 8(4): 293-332.

[30] 张慧, 张凡. 认知负荷理论综述[J]. 教育研究与实验, 1999(04): 45-47.

[31] 李爱梅, 颜亮, 王笑天, 等. 时间压力的双刃效应及其作用机制[J]. 心理科学进展, 2015, (09): 1627-1636.

[32] 李雪玮, 赵晓华, 黄利华, 等. 桥形标复杂度对驾驶人脑电认知特性的影响机理[J]. 西南交通大学学报, 2021, 56(05): 913-920.

[33] 周垚, 陈登凯, 谭晓雪, 等. 时间压力下驾驶舱界面视觉搜索绩效研究[J]. 东北大学学报(自然科学版), 2023, 44(10): 1514-1520.

[34] Merve Keskin et al. Exploring the Cognitive Load of Expert and Novice Map Users Using EEG and Eye Tracking[J]. ISPRS International Journal of Geo-Information, 2020, 9(7).

[35] Benguigui, Nicolas, Faure, et al. The effects of driving environment complexity and dual tasking on drivers' mental workload and eye blink behavior[J]. Transportation research, Part F. Traffic psychology and behaviour, 2016, 40F(Jul.): 78-90.

[36] AnmarkrudØ, Andresen A,Bråten I. Cognitive load and working memory in multimedia learning:Conceptual and measurement issues[J]. Educational Psychologist, 2019, 54(2): 61-83.

[37] 袁晓芳, 闫晶, 乔伟刚, 等. 社会助长效应对认知负荷影响的神经表征分析[J]. 中国安全生产科学技术, 2023, 19(07): 178-185.

[38] 郑霞忠, 王毓, 陈云, 等. 建筑施工安全培训信息呈现方式学习效果评价[J]. 中国安全科学学报, 2023, 33(05): 96-102.

[39] 徐丽丽, 栗继祖, 徐新华. 基于智能综采集控台人机系统的认知负荷研究[J]. 中国煤炭, 2023, 49(08): 54-64.

[40] 孟荣华, 李子奇, 吴正佳, 等. 基于装配任务与EEG功率信息特征的操作员认知负荷研究[J]. 机械设计, 2022, 39(11): 60-70.

[41] He Dengbo, Wang Ziquan, Khalil Elias B, et al. Classification of Driver Cognitive Load: Exploring the Benefits of Fusing Eye-Tracking and Physiological Measures[J]. Transportation Research Record, 2022, 2676(10): 670-681.

[42] 赵蕊. 智能化综采集控台操作员人因失误研究[D]. 辽宁: 辽宁工程技术大学, 2023.

[43] 李坤宸, 袁伟, 李舒欣, 等. 听觉-言语认知负荷下工作记忆容量对跟车行驶影响[J]. 交通运输系统工程与信息, 2024, 24(01): 93-105.

[44] 姜乃琪. 基于多源动态信息的飞行员认知负荷研究[D]. 江苏: 南京航空航天大学, 2023.

[45] 郑玲, 乔旭强, 倪涛, 等. 基于多维信息特征分析的驾驶人认知负荷研究[J]. 中国公路学报, 2021, 34(04): 240-250.

[46] Benedetti S, Pedrotti M, Minin L, et al. Driver Workload and Eye Blink Duration[J]. Transportatio Research Part F:Traffic Psychology and Behaviour, 2011, 14(3): 99-208.

[47] Rudin-Brown C M, Young K L, Patten C, et al. Driver Distraction in an Unusual Environment: Effects of Text-messaging in Tunnels[J]. Accident Analysis and Prevention, 2013, 50: 122-129.

[48] Caird J, Johnston K A, Willness C R, et al. The Effects of Texting on Driving Performance in a Driving Simulator:The Influence of Driver Age[J]. Accident Analysis and Prevention, 2014, 71: 311-318.

[49] Johan Engström, Mikael Ljung Aust, Matias Viström. Effects of Cognitive Load and Anticipation on Driver Responses to a Critical Traffic Event[J]. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 2018, 62(1): 1584-1588.

[50] Song C, Qu Z, Blumm N, et al. Limits of Predictability in Human Mobility[J]. ence, 2010, 327(5968): 1018-1021.

[51] Gonzalez M C, CA Hidalgo, Barabasi A L. Understanding individual human mobility patterns[J]. Nature, 2008, 453(7196): 779-782.

[52] 葛绍林, 叶剑, 何明祥. 基于深度森林的用户购买行为预测模型[J]. 计算机科学, 2019, 46(9): 190-194.

[53] 孙勇, 钟卓玲, 李璐. 基于班组安全氛围的建筑工人不安全行为预测[J]. 中国安全生产科学技术, 2019, (08): 156-161.

[54] D.A. Patel, K.N. Jha. Evaluation of construction projects based on the safe work behavior of co-employees through a neural network model[J]. Safety Science, 2016, 89: 240-248.

[55] Kahneman, D. Attention and effort[M]. Englewood Cliffs, NJ: Prentice-Hall, 1973.

[56] 黄希庭. 心理学导论[M]. 北京: 人民教育出版社, 2007.

[57] 康立新. 国内图式理论研究综述[J]. 河南社会科学, 2011, 19(04): 180-182.

[58] Paas F G W C, Merrinboer J J G V. The Efficiency of Instructional Conditions: An Approach to Combine Mental Effort and Performance Measures[J]. Human Factors The Journal of the Human Factors and Ergonomics Society, 1993, 35(4): 737-743.

[59] Meister D. Behavioral foundations of system development[J]. American Journal of Psychology, 1976, 90(2).

[60] Ryu K, Myung R. Evaluation of mental workload with a combined measure based on physiological indices during a dual task of tracking and mental arithmetic[J]. International Journal of Industrial Ergonomics, 2005(11): 35.

[61] 曹琦. 人的不安全行为分析及治理[J]. 人类工效学, 1998, 12(4): 36-38.

[62] 孙林岩. 人因工程[M]. 北京: 中国科学技术出版社, 2001.

[63] 陈红, 祁慧, 谭慧. 基于特征源与环境特征的中国煤矿重大事故研究[J]. 中国安全科学学报, 2005, 15(9): 33-38.

[64] 周刚, 程卫民. 人因失误与人不安全行为相关原理的分析与探讨[J]. 中国安全科学学报, 2008, 18(3): 10-14.

[65] Reason J. Human Error[M]. Cambridge: Cambridge university press, 1990.

[66] 刘轶松. 安全管理中人的不安全行为的探讨[J]. 西部探矿工程, 2005(06): 226-228.

[67] 曹庆仁. 浅析煤矿员工不安全行为的影响因素[J]. 矿业安全与环保, 2007, 33(6): 80-82.

[68] 孙勇, 钟卓玲, 李璐. 基于班组安全氛围的建筑工人不安全行为预测[J]. 中国安全生产科学技术, 2019, 15(8): 156-161.

[69] CHENG C, LEU S, CHENG Y, et al. Applying data mining techniques to explore factors contributing to occupational injuries in Taiwan's construction industry[J]. Accident Analysis & Prevention, 2012, 48: 214-222.

[70] 胡飞翔, 周建亮, 林欣燕, 等. 情绪对建筑工人不安全行为影响的认知试验[J]. 中国安全科学学报, 2023, 33(04): 75-83.

[71] Reimer B. Impact of cognitive task complexity on drivers’ visual tunneling[J]. Transportation Research Record, 2009, 2138(1): 13-19.

[72] 牛莉霞, 刘谋兴, 李乃文, 等. 工作倦怠、安全注意力与习惯性违章行为的关系[J]. 中国安全科学学报, 2016, 26(6): 19-23.

[73] 杨妍. 矿工疲劳与不安全行为的实验研究基于神经管理学的视角[D]. 西安: 西安科技大学, 2013.

[74] C.M. Privitera, L.W. Stark, Algorithms for defining visual regions-of-interest: comparison with eye fixations, IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(9): 970-982.

[75] Walter K, Bex P. Does Cognitive Load Affect Eye Movements/Oculomotor Behavior in Natural Scenes?[J]. 2021, 11(1)12405.

[76] Granholm E, Asarnow R F, Sarkin A J, et al. Papillary responses index cognitive resource limitations. Psychophysiology, 33, 457-461[J]. PubMed, 1996, 33(4): 457-461.

[77] Morimoto C H, Istance H, Spencer S N, et al. Proceedings of the Symposium on Eye Tracking Research and Applications[J]. 2008. 69-72.

[78] 陈伟炯, 冯祥, 李新, 等. 基于随机森林算法的叉车驾驶员情景意识水平分析[J]. 重庆交通大学学报(自然科学版), 2023, 42(10): 100-108.

[79] 向恬君. 山区公路超车及会车场景下驾驶人心生理及操作特性研究[D]. 重庆: 重庆交通大学, 2023.

[80] 陈发城, 鲁光泉, 林庆峰, 等. 有条件自动驾驶下驾驶人接管行为综述[J/OL]. 吉林大学学报(工学版): 1-16[2024-04-14]. https://link.cnki.net/urlid/22.1341.t.20240306.1055.001

[81] Stojmenova K, Sodnik J. Validation of Auditory Detection Response Task Method for Assessing the Attentional Effects of Cognitive Load[J]. Traffic Injury Prevention, 2018, 19(5): 495-500．

[82] 高梦圆. 考虑景观围合度的长直线公路驾驶行为特性研究[D]. 青岛: 青岛理工大学土木工程学院, 2023.

[83] 丁柏群, 闫艳红, 杨柳, 等. 基于模拟驾驶的借道左转驾驶人视觉特性研究[J]. 重庆交通大学学报(自然科学版), 2024, 43(03): 45-53.

[84] 张沥今, 魏夏琰, 陆嘉琦, 等. Lasso回归: 从解释到预测[J]. 心理科学进展, 2020, 28(10): 1777-1791.

[85] 杨俊闯, 赵超. K-Means聚类算法研究综述[J]. 计算机工程与应用, 2019, 55(23): 7-14+63.

[86] COVER T M, HART P E. Nearest neighbor pattern classification[J]. IEEE Transaction on Information The ory, 1967, 13(1): 21-27.

[87] 许子明, 牛一帆, 温旭云, 等. 基于脑电信号的认知负荷评估综述[J]. 航天医学与医学工程, 2021, 34(04): 339-348.

[88] 屠恩美, 杨杰. 半监督学习理论及其研究进展概述[J]. 上海交通大学学报, 2018, 52(10): 1280-1291.

[89] Breiman L. Random Forests[J]. Machine Learning, 2001, 45(1).

[90] 杨剑锋, 乔佩蕊, 李永梅, 等. 机器学习分类问题及算法研究综述[J]. 统计与决策, 2019, (06): 36-40.