摘要
冲击地压监测信息往往存在不准确、模糊甚至相互矛盾的问题,经典D-S理论在对高冲突信息进行融合时常常出现“Zadeh悖论”。采用D-S证据冲突概率平均加权融合算法进行处理,即将λ·f(A)替代证据冲突概率赋值函数f(A),有效解决了融合结果易失真的问题。将改进的D-S证据理论应用于冲击地压多源监测信息的融合预警,得到冲击概率δ_(hp),能够实现对高冲突信息的一致性描述。通过对冲击概率δ_(hp)进行实时修正,得到冲击态势δ_(rs),将冲击态势δ_(rs)和背景概率相结合,得到了冲击地压发生的概率增益指标G(A),有效消除了危险指数受预测对象时、空、强、范围及背景概率的影响,进一步提高了融合结果的准确性和可操作性。
The monitoring information of rock burst is often inaccurate,fuzzy and even contradictory,and the classical D-S theory appears"Zadeh Paradox"frequently when fusing high conflict information.The D-S evidence conflict probability average weighted fusion algorithm is used,that is,usingλ·f(A)replaces the evidence conflict probability assignment function f(A),which effectively solves the problem of easy distortion of fusion results.The improved D-S evidence theory was applied to the early warning of multi-source monitoring information of rock burst,the rock burst probabilityδ_(hp) was obtained,and the consistent description of high conflict information is realized.Through the real-time correction of the rock burst probabilityδ_(hp),the rock burst situationδ_(rs) was obtained.Combining the rock burst situationδ_(rs) with the background probability,the probability gain index G(A)of the occurrence of rock burst is obtained,which effectively eliminates the influence of the risk index affected by the time,space,strength,range and background probability of the prediction object,and further improves the accuracy and operability of the fusion results.
作者
夏永学
陆闯
冯美华
Xia Yongxue;Lu Chuang;Feng Meihua(CCTEG Coal Mining Research Institute Co.,Ltd.,Beijing 100013,P.R.China;Coal Mining and Designing Department,Tiandi Science and Technology Co.,Ltd.,Beijing 100013,P.R.China;State Key Laboratory of Coal Mining and Clean Utilization,China Coal Research Institute,Beijing 100013,P.R.China)
出处
《地下空间与工程学报》
CSCD
北大核心
2022年第4期1082-1088,共7页
Chinese Journal of Underground Space and Engineering
基金
国家自然科学基金项目(52174118)
中煤科工重点资助项目(2018-TD-ZD007)。
关键词
冲击地压
D-S证据理论
信息融合
预警模型
效能检验
rock burst
D-S evidence theory
information fusion
early warning model
performance test