摘要
我国煤炭所处的地质条件十分恶劣,随着矿井深度的增加,隧道异常发生的概率也在增加。通过引用无监督域自适应的方法来改善传统迁移学习的数据集昂贵和标签数据稀少的问题,采取一种基于对抗网络的域自适应方法,将地表的隧道异常样本迁移到矿井隧道的样本中去。通过实验数据显示,该方法比传统的神经网络算法提升了7.99%的准确率,取得了一定的成效。
The geological conditions of coal in my country are very bad.As the depth of the mine increases,the probability of tunnel disease also increases.By citing the method of unsupervised domain adaptation to improve the problems of expensive datasets and sparse labeled data in traditional transfer learning,a domain adaptation method based on adversarial network is adopted to transfer the abnormal samples of tunnels on the surface to the ones of mine tunnels sample.The experimental data shows that this method improves the accuracy by 7.99%compared with the traditional neural network algorithm,and has a certain effect.
作者
张洪明
许金星
ZHANG Hongming;XU Jinxing(Jiangsu Vocational College of Electronics and Information,Huaian 223003,China)
出处
《煤炭技术》
CAS
北大核心
2023年第5期159-161,共3页
Coal Technology
关键词
隧道异常
矿机隧道
域自适应
对抗网络
tunnel disease
miner tunnel
domain adaptation
adversarial networks
