The advancement of pre-trained language models (PLMs) has provided new avenues for addressing text classification challenges. This study investigates the applicability of PLMs in the categorization and automatic classification of short-text safety hazard information specifically within mining industry contexts. Leveraging the superior word embedding capabilities of encoder-based PLMs, the standardized hazard description data collected from mine safety supervision systems were vectorized while preserving semantic information. Utilizing the BERTopic model, the study successfully mined hazard category information, which was subsequently manually consolidated and labeled to form a standardized dataset for training classification models. A text classification framework based on both encoder and decoder models was designed, and the classification outcomes were compared with those from ensemble learning models constructed using Naive Bayes, XGBoost, TextCNN, etc. The results demonstrate that decoder-based PLMs exhibit superior classification accuracy and generalization capabilities for semantically complex safety hazard descriptions, compared to Non-PLMs and encoder-based PLMs. Additionally, the study concludes that selecting a classification model requires a comprehensive consideration of factors such as classification accuracy and training costs to achieve a balance between performance, efficiency, and cost. This research offers novel insights and methodologies for short-text classification tasks, particularly in the application of PLMs in mine safety management and hazard analysis, laying a foundation for subsequent related studies and further improvements in mine safety management practices.

Analyzing the mechanisms of accidents is essential for clarifying the accident evolution process, devising preventive measures, and achieving proactive accident management. To address the potential issues in existing accident causation theories, such as the unclear distinction between direct causes and incomplete accident evolution pathways in enterprise-level accident prevention analysis, this study systematically reviewed the elements involved in safety management activities and their interrelationships. We identified the central role of human factors in the accident evolution process and developed a full lifecycle evolution model for industrial accidents, which begins with hazard identification and follows a safety management logic as its primary framework. This model provides a clear pathway for constructing enterprise-level risk control lists and accident prevention schemes. The model’s effectiveness was validated through its application to China’s underground metal mining industry. Drawing on Chinese laws and regulations as well as accident investigation reports, this study identifies 11 common types of accidents in underground metal mines and maps their evolution pathways from a complex systems perspective. Quantitative data from 61 accident reports were used to pinpoint the core factors and critical pathways leading to these various accidents. The study also analyzes prevention strategies and proposes new countermeasures to control the propagation of accident risks. Practical applications of the model demonstrate that emphasizing human factors enhances the effectiveness and accuracy of enterprise-level accident analysis and risk management.

Investigating hidden hazards and implementing closed-loop management are essential strategies for accident prevention in the mining industry. This study tackles a key challenge in applying association rule mining to the development of hazard management plans for underground mines. The current approach mainly focuses on hazard description data, often underutilizing critical information such as hazard time and location. To address this, we integrate topic mining with association rule mining to uncover intrinsic association patterns among various attributes of mine safety hazards. Through a systematic analysis of standardized mining hazard attributes, five key analytical dimensions were identified: Hazard Type, Level, Time, Location, and Responsible Units. A topic mining model, utilizing the Biterm Topic Model, was constructed to reduce dimensionality and aggregate hazard description data. Evaluation indicators such as Standard Lift and Difference Degree were proposed, resulting in a multidimensional association rule mining model for mining safety hazards. In this research, 1387 valid rules were extracted based on hazard inspection data from an underground gold mine in China. The analysis revealed relatively strong associations between hazard location and hazard type, responsible unit, as well as hazard level, with association degrees of 1.934, 1.412, and 1.240, respectively. Additionally, 15 rules with a high degree of differentiation were identified to explore interesting correlations among different attributes. Based on this, corresponding control measures and improvement plans were developed for 19 locations. The results demonstrate that a multidimensional partition-based association rule mining approach for mining safety hazards can significantly enhance the specificity of safety training and improve the efficiency of safety hazard investigation.

Nowadays, the mining industry actively advocates and practices the concept of green and integrated exploitation to realize the sustainable development of resources with low-carbon emissions. The certain carbon quota for mining companies limits the production capacity and resource utilization efficiency. The integrated exploitation of multiple deposits could coordinate resource allocation and operation facilities, which would reduce capital expenditure and operating costs for the mining company from a systematic perspective. In this condition, some deposits located nearby could be treated as one entity to make plans and optimize. An optimization framework is proposed based on the analysis of the characteristics and advantages of integrated exploitation. A new mathematical programming model is presented to optimize production capacity and extracted ore grade for each deposit considering constraints of maximum and minimum mining capacity, extracted ore grade and concentrated ore grade requirement, and metal output target, which has a significant influence on the economic benefit and resource utilization rate for a mining company. The model is verified using the data collected from three deposits of a gold mining company in China to demonstrate its ability to optimize the allocation of production capacity and improve the technical and economic effect of mining under the limitation of carbon quota. The sensitivity analysis of some key parameters is carried out to generate a series of integrated exploitation schemes under different production and operation conditions, which is useful for the mining company to make decisions in different situations.