摘要
针对BP神经网络的不足,为提高概率积分法预计开采沉陷的准确性,采用遗传算法(GA)优化BP神经网络,建立了一种基于GA-BP神经网络的概率积分法参数预计模型。将多组地表观测站实测数据分为训练样本和检验样本,以工作面的7个地质采矿条件参数为输入集,5个概率积分法预计参数为输出集,通过GA优化的BP神经网络机器学习方法对训练样本进行训练,利用训练模型预计检验样本的概率积分法参数,并与观测站实测数据进行了对比分析。研究表明:对于不同地质采矿条件下的概率积分法参数进行预计时,GA-BP模型明显优于BP神经网络和偏最小二乘模型,平均相对误差最大为8.64%,预计精度可靠性较高。
Based on the shortcomings analysis of BP neural network,in order to improve the correctness in mining subsidence prediction by probability integral method,a probabilistic integration method parameter prediction model based on GABP neural network is established,which relied on the BP neural network principle optimized by genetic algorithm(GA).The measured data of surface observing stations are divided into training and testing samples.With the seven geological mining condition parameters as the input set,and the five prdeiction parameters of probability integral method are taken as the output set,training samples are trained by GA-optimized BP neural network machine learning method,and the parameters of probability integral methods for the test samples are predicted by the trained model,and compared with the measured data from the observation station.The study results indicate that the GA-BP model is superior to the BP neural network and the partial least squares model when predicting the parameters of the probability integral methods under different geological conditions,with the average relative error at 8.64%,and higher prediction accuracy.
作者
牛亚超
徐良骥
张坤
叶伟
张劲满
姜宝兴
Niu Yachao;Xu Liangji;Zhang Kun;Ye Wei;Zhang Jinman;Jiang Baoxing(School of Surveying and Mapping,Anhui University of Science and Technology,Huainan 232001,China;National Key Experiment of Mining Response and Disaster Prevention and Control in Deep Coal Mine,Huainan 232001,China)
出处
《金属矿山》
CAS
北大核心
2019年第10期93-100,共8页
Metal Mine
基金
国家自然科学基金项目(编号:41472323)
安徽省对外科技合作计划项目(编号:201904b11020015)
关键词
开采沉陷
概率积分法
遗传算法
BP神经网络
参数预计
Mining subsidence
Probability integral method
Genetic algorithm(GA)
BP neural network
Parameter prediction
