Quality Improvement of Cigarette Manufacturing Process Driven by Deep Learning: Application of Wasserstein Distance Algorithm

This paper aims to improve the quality monitoring level in the cigarette silk production, thereby proposing an innovative model based on a combination of Wasserstein distance and deep learning algorithms. This model combines Convolutional Neural Network (CNN), Bidirectional Gated Recurrent Unit (BiGRU), and attention mechanism to process the collected multi-dimensional process parameters and quality inspection data. The Wasserstein distance is used to calculate the differences between the distribution of process parameters in different time periods or batches, and it is input as a feature vector into the deep learning model. Finally, the performance of the model is experimentally evaluated. The results show that the proposed model significantly outperforms other models in key performance indicators such as accuracy, recall, and F1 value, with an accuracy of 96.08%. The recall and F1 values are 89.25% and 90.95%, respectively, and the prediction error is excellent, significantly reducing Mean Absolute Error (MAE) and Mean Relative Square Error (MRSE). The MAE and MRSE values of the proposed model reach 7.30 and 12.44, respectively. Therefore, the proposed model can improve the quality control level in the production process of cigarette silk production quality monitoring, providing solid data support and decision-making basis for production optimization.