Dual Attention Domain Adversarial Neural Network for Robust E-Nose Drift Compensation

Gas sensor drift is characterized by randomness and non-linearity, which is an unavoidable issue in electronic nose systems. To address this problem, a novel drift compensation method based on dual attention mechanisms and domain adversarial neural network (DA-DANN) is proposed in this paper. The core of this method lies in designing a feature extractor that includes parallel Position Attention Module and Channel Attention Module, aiming to adaptively capture the most discriminative spatial and channel features in the sensor array responses. Combined with the domain adversarial training strategy, the model can learn domain-invariant features between the source domain and the target domain, thereby effectively alleviating the data distribution discrepancy. In addition, the generalization ability of the model is further enhanced by introducing entropy minimization loss on the target domain. Experimental results show that compared with various baseline methods, the proposed model achieves a significant improvement in classification accuracy on public datasets, demonstrating its superiority and effectiveness in E-nose drift compensation tasks.