基于多任务学习模型的浅埋煤层大采高工作面顶板来压预测  

Roof weighting prediction of large mining height working face in shallow coal seam based on multi-task learning model

作　　者：张昆 贺广文 王青和 王泽 ZHANG Kun;HE Guangwen;WANG Qinghe;WANG Ze(China Coal Electric Co.,Ltd.,Beijing 101300,China;Kongzhuang Coal Mine,Shanghai Datun Energy Co.,Ltd.,Xuzhou 221600,China)

机构地区：[1]中煤电气有限公司,北京101300 [2]上海大屯能源股份有限公司孔庄煤矿,江苏徐州221600

出　　处：《陕西煤炭》2025年第4期119-124,共6页Shaanxi Coal

摘　　要：工作面顶板来压机理复杂,工作面倾向长度、埋深、顶板条件、推进速度、采高、基本顶厚度,直接顶厚度均对顶板来压有显著影响。分别对初次来压强度、初次来压步距、周期来压强度、周期来压步距及支架阻力做出预测,发现相比于单任务学习模型,多任务学习的预测效果更佳,拟合程度更高。该研究可以帮助工作人员及时做出调整,提前做出预防措施,减少顶板事故发生的可能性,保障安全生产。The roof weighting mechanism in the working face is complex,with factors such as the inclined length of the working face,buried depth,roof conditions,advancing speed,mining height,basic roof thickness,and direct roof thickness all significantly affecting roof weighting.The strengths and steps of both the initial weighting and periodic weighting,as well as support resistance,are predicted,respectively.It has been found that multi-task learning yields a better prediction effect and a higher degree of fit compared to single-task learning models.Our study can assist staff in making timely adjustments,taking preventive measures in advance,reducing the likelihood of roof accidents,and ensuring safe production.

关 键 词：顶板来压 来压强度 周期来压步距 模型预测 

分 类 号：TD323[矿业工程—矿井建设]