采煤机司机作为煤矿企业的核心工种，其主要负责采煤机操作运行工作，完成本岗割煤任务。由于复杂的工作环境和单调重复的工作内容，采煤机司机容易处于过高或过低的心理负荷状态，不良的心理负荷会通过异常的操作状态表征出来，进而产生不安全行为。因此，探究采煤机司机作业过程中心理负荷的变化，对采煤机司机的不良心理负荷进行及时准确地识别并采取应对措施，从而预防和控制煤矿人因事故的发生尤为重要。然而，目前对采煤机司机心理负荷的研究仍较缺乏。鉴于此，本研究采用便携式功能近红外设备同步采集被试的功能性近红外光谱（functional near-infrared spectroscopy，fNIRS）信号、行为学数据和主观心理负荷量表得分，利用K-means聚类法对心理负荷进行分级，并基于机器学习算法建立采煤机司机心理负荷识别模型，对其作业过程中的不良心理负荷进行识别，提出管控措施，进而完善煤矿企业安全管理。主要研究内容和结论如下：

（1）设计并完成了采煤机司机心理负荷诱发实验。分别采集被试在警惕操作任务和风险图片判断任务下的近红外指标（HbO）和平均反应时间、正确率等行为学指标，使用美国航天局心理负荷（NASA-TLX）量表获取被试的主观心理负荷得分。使用SPSS27.0对行为学指标和NASA-TLX量表得分进行统计分析，验证本研究心理负荷诱发实验的科学性、有效性与适用性。

（2）分析氧合血红蛋白浓度（HbO）和行为数据的变化。将任务流程按照时间序列进行分段，采用相关信号改善（CBSI）算法对近红外数据进行处理，分析fNIRS数据与大脑活动区域的关系。结果表明在被试大脑前额叶感兴趣区域（ROI），任务进行的前后不同阶段，被试HbO浓度均值呈现显著差异，反映出被试在任务进行过程中心理负荷的变化。通过SPSS27.0对被试行为数据中的平均反应时间和正确率进行单因素方差分析和成对比较分析，发现随着工作时间的延长，被试的反应时间和错误率均显著增加。

（3）对心理负荷程度进行分级，建立基于fNIRS的采煤机司机心理负荷识别模型。对实验所获得的行为学数据、NASA-TLX得分和近红外指标进行K-means聚类分析，根据聚类结果，将矿工的心理负荷等级划分为低、中、高三级。优选采煤机司机心理负荷表征指标作为输入向量，心理负荷等级作为输出向量，分别应用K近邻、支持向量机和朴素贝叶斯三种分类算法构建采煤机司机心理负荷识别模型。通过计算并综合对比三种识别算法的总体识别正确率、精确率、召回率和F1值，得出K近邻算法下的识别模型效果最优。

综上，本研究实现了通过采集fNIRS信号来监测采煤机司机作业过程中的心理负荷状态，通过对心理负荷进行分级并构建识别模型，来识别采煤机司机作业过程中的不良心理负荷，并根据研究结果提出防控对策及建议，从而为煤矿企业管理者采取有效的、有针对性的安全管理举措提供理论依据与技术支撑。

As the core type of work in coal mining enterprises, the shearer driver is mainly responsible for the operation of the shearer and the completion of the coal cutting task. Due to the complex working environment and monotonous and repetitive work content, the shearer driver is easy to be in a state of too high or too low psychological load. The bad psychological load will be characterized by abnormal operating conditions, which will lead to unsafe behavior. Therefore, it is particularly important to explore the changes in the psychological load of the shearer driver during the operation process, identify the adverse psychological load of the shearer driver in a timely and accurate manner and take countermeasures, so as to prevent and control the occurrence of human accidents in coal mines. However, there is still a lack of research on the psychological load of shearer drivers. In view of this, this study used a portable functional near-infrared device to synchronously collect the functional near-infrared spectroscopy (fNIRS) signals, behavioral data, and subjective psychological load scale scores of the subjects. The K-means clustering method was used to classify the psychological load. Based on the machine learning algorithm, a shearer driver 's psychological load identification model was established to identify the adverse psychological load during its operation, and control measures were proposed to improve the safety management of coal mine enterprises. The main research contents and conclusions are as follows :

( 1 ) Designed and completed the shearer driver psychological load induction experiment. The near-infrared index (HbO), average reaction time, correct rate and other behavioral indicators of the subjects under the vigilance operation task and the risk picture judgment task were collected respectively. The NASA-TLX scale was used to obtain the subjective psychological load score of the subjects. SPSS27.0 was used to analyze the behavioral indicators and NASA-TLX scale scores to verify the scientificity, effectiveness and applicability of the psychological load induction experiment in this study.

( 2 ) The changes of oxyhemoglobin concentration (HbO) and behavioral data were analyzed. The task flow was segmented according to the time series, and the correlation signal improvement (CBSI) algorithm was used to process the near-infrared data to analyze the relationship between fNIRS data and brain activity regions. The results showed that in the prefrontal region of interest (ROI) of the subjects ' brain, the mean HbO concentration of the subjects showed significant differences at different stages before and after the task, reflecting the changes of the subjects ' psychological load during the task. Through SPSS27.0, the average reaction time and correct rate in the behavior data of the subjects were analyzed by one-way ANOVA and paired comparative analysis. It was found that the reaction time and error rate of the subjects increased significantly with the extension of working time.

( 3 ) The degree of psychological load is classified, and the identification model of shearer driver 's psychological load based on fNIRS is established. K-means clustering analysis was carried out on the behavioral data, NASA-TLX score and near-infrared index obtained from the experiment. According to the clustering results, the psychological load level of miners was divided into three levels : low, medium and high. The psychological load characterization index of the shearer driver is selected as the input vector, and the psychological load level is used as the output vector. The three classification algorithms of K-nearest neighbor, support vector machine and naive Bayes are used to construct the psychological load identification model of the shearer driver. By calculating and comprehensively comparing the overall recognition accuracy, accuracy, recall rate and F1 value of the three recognition algorithms, it is concluded that the recognition model under the K-nearest neighbor algorithm has the best effect.

In summary, this study realizes the monitoring of the psychological load state of the shearer driver during operation by collecting fNIRS signals. By classifying the psychological load and constructing the identification model, the adverse psychological load of the shearer driver during operation is identified, and the prevention and control countermeasures and suggestions are put forward according to the research results, so as to provide theoretical basis and technical support for the managers of coal mine enterprises to take effective and targeted safety management measures.