煤矿事故中由于煤矿工人警觉性水平下降导致工作失误和不安全行为的案例不在少数，煤矿井下工作环境恶劣、复杂更易导致矿工注意力水平等的下降，因此研究不同环境湿度对警觉性的影响具有重要意义。为了深入研究湿度对矿工警觉性的影响机理，本文基于便携式近红外技术（fNIRS，functional Near-Infrared Spectroscopy）展开对矿工警觉性水平的研究，对实验过程中不同阶段的fNIRS数据、行为数据进行分析，并基于分析结果进行警觉性水平分类，在此基础上构建遗传算法优化的BP神经网络（GA-BP神经网络）进行矿工警觉性水平预测和识别，达到减少因警觉性水平下降导致的矿工在工作任务中不安全行为的产生，为煤矿企业的安全管理提出一种新角度。

本实验邀请了40名被试，模仿真实的煤矿井下作业场景构建实验方案，设置不同环境湿度条件，让被试进行运动以模拟煤矿工人作业情况；完成工作任务模拟后，依据精神运动警觉性任务范式测量并采集被试fNIRS数据，同时使用E-Prime软件收集行为数据，如反应时间、正确率等。用平均反应时间指标来表征警觉性水平并通过层次聚类算法进行分类，最终将警觉性水平分为活跃、正常、轻微下降、严重下降。通过对27个通道的激活β值进行独立样本t检验，分析结果发现不同湿度环境中通道3、6、9、15、27共五个通道具有显著差异。根据解剖标定体系将实验中的27个通道划分为8个感兴趣脑区，通过对不同脑区的信号激活和脑血氧浓度进行分析，结果发现高湿度环境中信号显著激活的区域范围较大，尤其集中于承担大脑认知、记忆以及决策功能的背外侧前额叶区和额极区。通过含氧血红蛋白浓度反映各脑区之间功能连接性，结果发现高湿度环境下进行工作任务需要脑网络之间的连接更协调、高效，验证了环境湿度基于生理代谢和认知负荷对警觉性产生影响。最后选择实验任务中fNIRS激活显著的脑区共14个通道不同时段的含氧血红蛋白浓度指标，结合平均反应时间作为输入数据，选择4个警觉性等级作为输出指标，构建了遗传算法优化的BP神经网络。之后通过精准率、召回率、f1-score评价指标计算分析，结果表明本文所构建的GA-BP神经网络预测模型对警觉性状态识别来说具有较高的精准性和良好的容错性。

本文研究了不同湿度环境对矿工警觉性水平的影响，进一步从脑活动特征揭示环境湿度对警觉性的影响机理及规律特点，为煤矿井下工作人员的警觉性水平检测和状态管理提供了科学依据，最终达到减少不安全行为发生的目的。

In coal mine accidents, there are many cases of work errors and unsafe behaviors caused by the decline of the miners' vigilance level. The working environment of coal mine underground is harsh and complex, which is more likely to lead to the decline of miners' attention level. Therefore, it is of great significance to study the influence of different environmental humidity on vigilance. In order to further study the mechanism of humidity's influence on miners' vigilance, the thesis carries out the research on miners' vigilance level based on portable near infrared technology (fNIRS, functional Near-Infrared Spectroscopy). The fNIRS data and behavior data of people in different experimental stages are analyzed, and the vigilance level is classified based on the analysis results. On this basis, the BP neural network optimized by genetic algorithm (GA-BP neural network) is constructed to predict and identify the miners' vigilance level, so as to reduce the occurrence of unsafe behaviors in the work task caused by the decline of the miners' vigilance level, and provides a new perspective for the safety management of coal mining enterprises.

This experiment invites 40 participants, creating an experimental design that simulated real coal mine underground working conditions by setting various environmental humidity levels. Participants were engaged in physical activities to mimic the working conditions of coal miners. Following the completion of the simulated work tasks, participants' functional Near-Infrared Spectroscopy (fNIRS) data were measured and collected based on the psychomotor vigilance task (PVT) paradigm. Behavioral data, such as reaction times and accuracy rates, were also gathered using E-Prime software. The average reaction time was used as an indicator to characterize the level of vigilance and was classified through hierarchical clustering analysis into four levels: active, normal, mildly decreased, and severely decreased. An independent sample t-test on the activation values β of 27 channels revealed significant differences in channels 3, 6, 9, 15, and 27 across different humidity environments. Based on the anatomical labeling system, the 27 channels in the experiment were divided into 8 regions of interest (ROI) within the brain. Analysis of signal activation and cerebral oxygenation levels in different brain regions indicated a broader range of significant signal activation in high humidity environments, particularly concentrated in the dorsolateral prefrontal cortex and frontal operculum areas, which are responsible for cognitive, memory, and decision-making functions. The connectivity between brain networks, reflected by oxyhemoglobin concentration, indicated that working tasks in high humidity environments require more coordinated and efficient connections between brain networks, confirming the impact of environmental humidity on vigilance through physiological metabolism and cognitive load. Finally, oxygenated hemoglobin concentration indexes of 14 channels in the brain regions with significant fNIRS activation were selected for different periods in the experimental task. Combined with the average reaction time as input data, four alertness levels were selected as output indexes, and a BP neural network optimized by genetic algorithm was constructed. Subsequent analysis based on precision, recall, and f1-score metrics demonstrated that the GA-BP neural network model constructed in this study has high accuracy and robust fault tolerance in recognizing vigilance states.

The thesis verified the influence of different humidity environments on the alertness level of miners, and further revealed the influence mechanism and regular characteristics of environmental humidity on alertness from the characteristics of brain activity, which provides a scientific basis for the detection and state management of the alertness level of coal miners, and ultimately achieves the purpose of reducing unsafe behaviors.