A Cigarette Loose End Prediction Approach Based on Physics-Informed Neural Networks

The cigarette loose end is one of the core indicators for quality control in cigarette production processes. The occurrence of cigarettes loose end directly affects the quality of cigarettes, thereby affecting the consumer’s smoking experience. This paper takes the cigarette loose end as the research object, collects multidimensional cigarette process parameter data, constructs a physical information loss term using the physical prior knowledge of the rolling process, and establishes a cigarette loose end prediction approach based on physical information neural network. The results indicate that the proposed approach can achieve accurate prediction of cigarette loose end.