摘要
【目的】冲击地压作为煤矿开采中的重大动力灾害,其精准预测对矿山安全生产至关重要。BP模型广泛应用于地灾预测,该模型存在收敛速度慢及全局最优解收敛不确定性等缺点。为了准确对冲击地压灾害进行预测,本研究提出一种基于非线性海洋捕食者算法优化的BP神经网络模型,用于煤样冲击倾向性预测。【方法】通过系统筛选动态破坏时间、弹性能量指数、冲击能量指数和单轴抗压强度4项关键指标,构建包含127组煤样的数据集。数据经归一化预处理后,按7∶3比例划分为训练集与测试集。【结果】为验证模型性能,同步采用鲸鱼优化算法、灰狼优化算法、粒子群算法、人工蜂群算法、标准海洋捕食者算法及传统BP模型进行对比实验。对比传统BP模型、MPA-BP模型,NMPA-BP模型在解决BP算法收敛速度慢及全局最优解收敛不确定性方面具有显著优势,其预测准确率达94.9%。【结论】在6项工程实例中,预测结果与实际风险等级完全吻合,该模型在煤样冲击倾向性预测中的有效性和实用性。
[Purposes]As a major dynamic hazard in coal mining,accurate prediction of rockburst is essential for ensuring mine safety.While the Back Propagation(BP)neural network model has been widely adopted for geological disaster prediction,it suffers from limitations such as slow convergence and uncertainty in attaining the global optimum.To improve the accuracy of rockburst prediction,this study proposes a BP neural network model optimized by an improved Nonlinear Marine Predators Algorithm(NMPA)for assessing coal burst propensity.[Methods]Through systematic evaluation,four key indicators were selected as model inputs:dynamic failure time,elastic energy index,impact energy index,and uniaxial compressive strength.A comprehensive dataset comprising 127 coal samples was established.After normalization preprocessing,the dataset was partitioned into training and testing subsets at a 7∶3 ratio.[Findings]To rigorously evaluate model performance,comparative analyses were conducted against five optimization algorithms:Whale Optimization Algorithm,Grey Wolf Optimizer,Particle Swarm Optimization,Artificial Bee Colony,standard Marine Predators Algorithm,and the conventional BP model.The proposed NMPA-BP model demonstrated superior performance in addressing the convergence limitations of traditional BP networks,achieving a prediction accuracy of 94.9%.[Conclusions]Validation across six engineering case studies revealed perfect alignment between predicted and actual risk classifications,confirming the model's reliability and practical applicability for coal burst propensity assessment.The findings provide a robust methodological framework for rockburst early warning systems in underground coal mines.
作者
尹胜
廖华江
刘清
王体富
陈碧松
叶洪盛
YIN Sheng;LIAO Huajiang;LIU Qing;WANG Tifu;CHEN Bisong;YE Hongsheng(Zunyi Aluminum Co.,Ltd.,Zunyi 563000,China)
出处
《河南科技》
2026年第1期48-55,共8页
Henan Science and Technology
