基于增强卷积神经网络的电子鼻长期漂移抑制方法

为解决电子鼻传感器阵列中的漂移问题,提出了一种增强卷积神经网络的长期漂移抑制方法。首先,通过结合历史数据的方式进行数据库扩增,起到了数据增强的效果;然后,使用增量补偿模块结合增量学习思维进行网络训练,起到了模型增强的效果;最后,分别使用公开数据集和实测数据集来验证模型的漂移抑制效果。实验结果表明:增强卷积神经网络算法的抑制效果较传统卷积神经网络、机器学习算法有较大提升,精度提高幅度为10%～20%,精度波动在1%范围内,具有较好的鲁棒性,验证了增强卷积神经网络算法在电子鼻漂移抑制中是稳健有效的,同时也从算法层面对电子鼻的漂移抑制提供了思路。

Long-term drift suppression method for electronic nose based on the augmented convolutional neural network

To solve the drift problem in the electronic nose sensor array, a long-term drift suppression method with the augmented convolutional neural network is proposed. First, by combining historical data to expand the database, it has the effectiveness of data enhancement. Then, the incremental compensation module is used for network training to enhance the entire network performance. Finally, public dataset and measured dataset are utilized to evaluate the drift suppression performance, respectively. Compared with the traditional convolutional neural network and machine learning algorithms, experimental results show that the proposed augmented convolutional neural network ( ACNN) has great accuracy increase about 10% - 20% , and the accuracy fluctuation of 1% is good robustness, which verified that the augmented convolutional neural network is robust and effective in the suppression of electron nose drift, at the same time, also provides ideas for the drift suppression of electronic nose from the algorithm level.