Defects Recognition in Filament Package Based on Improved Neural Network
The chemical fiber industry is a traditional pillar industry in China.In the process of filament production,the environment and production technology and other factors can easily lead to defects in the appearance of filament winding,which affects weaving efficiency and filament quality to some extent.Most of the appearance of filament winding is identified manually,with the problem of slow speed and low accuracy.Thus it proposes the SCNet model,a ConvNeXt convolutional neural network model that integrates the attention mechanism SimAm,which increases the model's ability to extract defect features while maintaining its lightness.During the experimental process,data augmentation and transfer learning are used to enrich the diversity of samples and accelerate the convergence speed of the model.The experiment shows that the SCNet model has a recognition accuracy of 96.6％for the appearance of defective filament package.The model has good classification performance for normal and three common types of filament package appearance defects with good universality.
Appearance of Filament Roll PackagingDefect DetectionDeep LearningConvNeXtSimAmSCNet
万方数据
维普
