矿区开采沉陷预计的改进BP神经网络模型

发布时间：2018-04-24 13:55

本文选题：开采沉陷 + BP神经网络模型　；参考：《金属矿山》2017年04期

【摘要】：为精确预计锦界矿某工作面开采沉陷,首先结合该工作面的地质资料、采掘工作平面图及孔柱状图,采用FLAC3D软件建立了该工作面开采沉陷仿真模型,得到工作面推进100、300、500、700 m时的开采沉陷数据;其次基于该类数据对BP神经网络预计模型进行训练和验证,建立沉陷数据与工作面推进距离的非线性关联;然后用粒子群优化算法(Particle swarm optimization,PSO)对BP神经网络模型的结构参数和连接权值阈值进行优化,并引入遗传算法(Genetic algorithm,GA)中的自适应变异因子以一定概率初始化部分变量,以解决PSO算法易陷入局部最优解的问题,避免BP神经网络模型易陷入局部最小值、训练收敛速率低以及PSO算法易早熟收敛等问题。分别采用BP神经网络模型、PSO-BP神经网络模型以及所提模型进行试验对比,并引入偏差平方和(Sum of squares for total,SST)对各模型的预计精度进行评价,研究表明:在工作面分别推进100,300,500 m的情况下,BP神经网络模型的SST值分别为0.056,0.062,0.066,PSO-BP神经网络模型的SST值分别为0.049,0.054,0.048,所提模型的SST值分别为0.028,0.026,0.031,明显小于前两者,表明该模型有助于提高矿区开采沉陷预计精度,有一定的实用价值。
[Abstract]:In order to accurately predict the mining subsidence of a working face in Jinjie Coal Mine, a simulation model of mining subsidence is established by using FLAC3D software, combining with the geological data of the working face, the excavation plan and the hole histogram. The mining subsidence data are obtained when the working face is advancing 100300500700 m. Secondly, the BP neural network prediction model is trained and verified based on this kind of data, and the nonlinear correlation between the subsidence data and the working face advance distance is established. Then, the particle swarm optimization algorithm (PSO) is used to optimize the structural parameters and the threshold of connection weights of BP neural network model, and the adaptive variation factor in genetic algorithm GAA is introduced to initialize some variables with a certain probability. In order to solve the problem that PSO algorithm is easy to fall into local optimal solution, the BP neural network model is easy to fall into local minimum, the rate of training convergence is low, and PSO algorithm is easy to converge prematurely. The BP neural network model, PSO-BP neural network model, and the proposed model are used to compare the experimental results, and the sum of squares squared of squares for total SSTs are introduced to evaluate the predicted accuracy of each model. The results show that the SST value of BP neural network model is 0.056 0. 062% and the SST value of PSO-BP neural network model is 0. 049 0. 054 卤0. 048, and the SST value of the proposed model is 0. 028 ~ 0. 026 ~ 0. 031, which is obviously smaller than that of the former two models. It shows that the model is helpful to improve the prediction accuracy of mining subsidence and has certain practical value.
【作者单位】：北京电子科技职业学院电信工程学院;解放军理工大学国防工程学院;爆炸冲击防灾减灾国家重点实验室;解放军61206部队;
【分类号】：TD327;TP183

【相似文献】