Intelligent Technology Center, CHN Energy Shendong Coal Group Co., Ltd.
Tiandi(Changzhou) Automation Co., Ltd.
School of Physical Education,China University of Mining and Technology

在煤炭开采过程中混入的异物可能会导致输送带连接处堵塞甚至输送带撕裂等事故,现有的机器学习算法大多采用监督学习的方式自动识别物品类别,而在真实工矿场景下,异常样本稀缺,易导致建模数据集存在严重的样本分布不平衡且显著特征丢失的问题。针对上述问题,提出了一种基于双注意力生成对抗网络(DA−GANomaly)的煤流异物智能检测方法。该方法采用半监督学习的方式,仅需要正常样本完成异物检测模型训练,有效解决了因样本分布不平衡造成的识别精度低、鲁棒性差的问题;在Skip−GANomaly的基础上引入双注意力机制,增强了编码器与解码器之间的信息交流,以抑制无关特征和噪声,同时突出有利于区分异常样本的感兴趣特征,进一步提高模型分类的准确性。实验结果表明:DA−GANomaly模型的分类精确率为79.5%,召回率为83.2%,精确率−召回率曲线下面积(AUPRC)为85.1%;与AnoGAN等5种经典异常检测模型相比,DA−GANomaly模型的综合性能最佳。

Foreign objects mixed in during coal mining may cause accidents such as blockage or even tearingof conveyor belt connections. Most existing machine learning algorithms for coal flow foreign objects usesupervised learning to automatically recoginze item categories. However, in real industrial and mining scenarios,the scarcity of abnormal samples leads to problems of serious imbalanced sample distribution and significantfeatures lost in the modeling dataset. In order to solve the above problems, a coal flow foreign object intelligentdetection method based on dual-attention Skip-GANomaly (DA-GANomaly) is proposed. This method adopts asemi supervised learning approach, which only requires normal samples to complete the training of the foreignobject detection model, effectively solving the problems of low recognition accuracy and poor robustness causedby imbalanced sample distribution. On the basis of Skip-GANomaly, a dual attention mechanism is introduced toenhance the information exchange between the encoder and decoder and suppress irrelevant features and noise. It highlights the interesting features that are conducive to distinguishing abnormal samples, and further improves theaccuracy of model classification. The experimental results show that the classification accuracy of the DAGANomaly model is 79.5%, the recall rate is 83.2%, and the area under the precision recall curve (AUPRC) is85.1%. Compared with 5 classic anomaly detection models such as AnoGAN, the DA-GANomaly model has thebest overall performance.

天地(常州)自动化股份有限公司科研项目(2022FY0009)

曹正远,蒋伟,方成辉. 基于双注意力生成对抗网络的煤流异物智能检测方法[J]. 工矿自动化,2023,49(12):56-62.

CAO Zhengyuan, JIANG Wei, FANG Chenghui. Intelligent detection method for coal flow foreign objects based on dual attentiongenerative adversarial network[J]. Journal of Mine Automation,2023,49(12):56-62.