随着矿井开采深度不断增加，导致采煤工的工作环境越来越恶劣，高温的问题及采煤工的职业健康问题愈加突出。并下的不良环境因素会引起采煤工的心理和生理变化，从而影响采煤工的身心状态，威胁煤矿安全生产。采煤工不安全行为是导致煤矿生产事故的主要原因，采煤工的状态差是发生不安全行为的前提，因此探究环境温度与采煤工不安全状态之间的关系，实现采煤工不安全状态识别预警，能够从根本上降低采煤工不安全行为对安全生产的不利影响。本文针对温度因素,采用主观量表及生理参数测量法开展了温度与采煤工不安全状态的影响实验研究。

首先，明晰采煤工不安全状态的影响因素、形成原因及测量指标,并提出研究假设。本文综合分析了采煤工的工作环境、工艺流程及高温环境形成原因，对温度与采煤工不安全状态的关系进行了分析，提出了影响路径及研究假设，确定了研究温度对采煤工不安全状态影响的主、客观测量指标，包括生理指标、心理指标及反应能力、风险识别能力指标。

然后，设计并开展了温度对采煤工不安全状态影响的实验，验证假设并建立了采煤工不安全状态表征指标集。实验记录了 35 名被试的主、客观测量指标数据，采用高通、低通滤波器对实验所得的生理信号数据进行了预处理，去除生理信号中的噪音干扰;运用 SPSS 26.0 软件对实验对象的心理状态特征指标、生理状态特征指标、反应能力特征指标、风险识别能力特征指标数据进行差异分析，明确了高温环境对采煤工不安全状态相关指标的影响，验证了研究假设;并基于灰色关联理论及主成分分析法对主、客观测量指标进行筛选、降维，从而建立采煤工不安全状态表征指标集,共7个综合表征指标。

最后，建立了基于温度的采煤工不安全状态预警模型。根据7个采煤工不安全状态综合表征指标，运用 K-means聚类法对采煤工的不安全状态进行分级;在此基础上比较XGBoost、BP神经网络、朴素贝叶斯三种分类算法对采煤工不安全状态的识别效果，结果显示 BP 神经网络算法模型的识别效果最好，说明该模型能够较好地识别采煤工不安全状态，因此选择 BP神经网络算法建立了采煤工不安全状态预警模型。

结果表明，在高温环境下采煤工感觉不适，高温对采煤工的心理状态、生理状态反应能力及风险识别能力均产生显著影响，随着温度升高，采煤工的安全状态星下降趋势。建立的不安全状态表征指标集，能够表征采煤工的不安全状态,将其分为安全状态、较不安全状态、不安全状态三个等级;建立的采煤工不安全状态识别及预警模型，能够较好地识别预警采煤工的不安全状态，将预警等级分为安全无预警、中度预警、重度预警，并提出了相应的对策措施，为煤炭企业开展安全管理、减少煤矿事故的发生及后续采煤工不安全状态预警平台的搭建提供理论依据。

With the increasing depth of mining, the working environment of coal miners is gettingworse and worse, and the problems of high temperature and occupational health of coal minersare becoming more and more prominent, The bad environmental factors in underground willcause the psychological and physiological changes of coal miners, which will affect the physicaland mental state of coal miners and threaten the safety of coal mine production. Unsafe behaviorof coal miners is the main cause of coal mine production accidents, and poor state of coal minersis the premise of unsafe behavior. Therefore, exploring the relationship between environmentaltemperature and unsafe state of coal miners and realizing identification and early warning ofunsafe state of coal miners can fundamentally reduce the adverse impact of unsafe behavior ofcoal miners on safety production. In this paper, the influence of temperature on the unsafe stateof coal miners is studied by using subjective scale and physiological parameter measurement.

Firstly, the influencing factors, formation reasons and measurement indicators of the unsafestate of coal miners are clarified, and the research hypothesis is put forward. This papercomprehensively analyzes the working environment, process flow and causes of hightemperature environment of coal miners, analyzes the relationship between temperature andunsafe state of coal miners, puts forward the influence path and research hypothesis, anddetermines the main and objective measurement indicators of the influence of temperature onunsafe state of coal miners, including physiological indicators, psychological indicators, reactionability and risk recognition ability indicators.

Then, an experiment on the influence of temperature on the unsafe state of coal workers isdesigned and carried out to verify the hypothesis and establish the index set of the unsafe stateof coal workers. The subjective and objective measurement data of 35 subjects were recorded inthe experiment. High-pass and low-pass filters were used to preprocess the physiological signaldata to remove the noise interference in the physiological signal. SPSS 26.0 software was used to make difference analysis on the data of psychological state characteristic indicators,physiological state characteristic indicators, reaction ability characteristic indicators and riskrecognition ability characteristic indicators of the experimental subjects, and the influence ofhigh temperature environment on the relevant indicators of the unsafe state of coal miners wasclarified, and the research hypothesis was verified. Based on grey correlation theory and principalcomponent analysis method, the main and objective measurement indicators are screened anddimensionality is reduced, in order to establish a representation index set of unsafe state of coalminers, with a total of7 comprehensive indicators.

Finally, the early-warning model of coal miner's unsafe state based on temperature isestablished, According to 7 comprehensive indicators of unsafe state of coal miners, K-meansclustering method is used to grade the unsafe state of coal miners. On this basis, the recognitioneffect of Extreme Gradient Boosting, Back Propagation Neural Network and Naive BayesClassification Algorithms on the unsafe state of coal miners is compared. The results show thatthe recognition effect of Back Propagation Neural Network Model is the best, indicating that thismodel can identify the unsafe state of coal miners well. Back Propagation Neural NetworkAlgorithm is selected to establish the unsafe state early warning model of coal miners.

The results show that coal workers feel uncomfortable obviously in high temperatureenvironment, and high temperature has a significant impact on the psychological state.physiological state, reaction ability and risk recognition ability ofcoal workers. With the increaseof temperature, the safety state of coal workers shows a downward trend. The established indexset of unsafe state can represent the unsafe state of coal miners, which can be divided into threelevels: safe state, less safe state and unsafe state. The established model of unsafe staterecognition and early waring for coal miners can identify and early warn the unsafe state of coalminers well, and the early waring levels are divided into safety no early warning, moderateearly waring and severe early warning, and corresponding countermeasures are proposedwhich provides a theoretical basis for coal enterprises to carry out safety management, reducethe occurrence of coal mine accidents and subsequently build an unsafe state early warningplatform for coal miners.