为提升违章行为管理的有效性和针对性，引入行为安全“2-4”模型（24Model）和自组织映射网络（SOM）研究违章行为的分布特征；利用24Model分析现场违章行为，构建违章行为分类体系；通过SOM对违章行为特征数据进行训练和学习，结合k-means算法分析部门的聚类特征，生成可视化拓扑结构，识别出关键违章行为。结果表明：得到的三级违章行为分类体系，识别出了违章行为数量较多的部门及其高频发违章行为；根据三级违章行为映射图，“未合理布置工艺设备及作业环境”、“未有效维护设备和设施”、“脱岗或离岗”等是该煤矿整体的高频发违章行为，广泛存在于多数部门中；根据二级违章行为映射图，“不安全操作”数量最高，主要为体现为“未按规程作业或作业不到位”，多发于生产技术部、钻机队；对关键违章行为及相应部门采取针对性措施，能有效降低因不安全行为导致事故发生的可能性。

To enhance the effectiveness and specificity of violation behavior management, the behavior security “2-4” model (24Model) and self-organizing mapping network (SOM) are introduced to study the distribution characteristics of violation behavior. The 24Model is used to analyze on-site violations and construct a violation classification system; the violation feature data is trained and learned through SOM, and the clustering features of the department are analyzed in conjunction with the k-means algorithm to generate a visual topology that identifies the key violations. The results indicate the establishment of a three-level classification system for violation behaviors; the study identifies departments with a high frequency of violations and their specific high-frequency violation behaviors; according to the three-level mapping of violations, “failure to reasonably arrange process equipment and working environment”, “failure to effectively maintain equipment and facilities”, “absence from work or duty” are high-frequency violations across the entire mine, which are widely found in most of the mines. According to the mapping of secondary violations, the number of “unsafe operation” is the highest, mainly reflecting “failure to operate according to regulations or operation not in place”, which mostly occurs in the production technology department and the drilling rig team. Adopting targeted measures for key violations and corresponding departments can effectively reduce the possibility of accidents caused by unsafe act.