Research On Quality Inspection Algorithm Of Rare Earth Magnetic Materials Based On Machine Vision

Rare earth is an important non renewable strategic resource.Rare earth magnetic materials are the most widely used and largest branch of rare earth materials,accounting for more than 60% of the application of new rare earth materials.With the intensification of global market competition,higher requirements have been put forward for the production of rare earth magnetic materials in terms of improving product quality,reducing resource consumption,reducing production costs,and improving service levels.The production of rare earth magnetic materials is characterized by multiple processes,multiple product types,multiple production equipment,multiple process parameters,and strong coupling.Traditional production management and control methods cannot better manage and utilize the data generated during the production process,and at the same time,there is a slight lag in production intelligence,and real-time monitoring and decision-making of data is not possible.Based on the research of 5G technology and digital twinning technology,this paper selects the molding production process in the production of rare earth magnetic materials as the main research object,aiming at the quality inspection issues and production data collection issues existing in the molding production process of rare earth magnetic materials.This article has completed the digitization and automation transformation of molding production equipment,and developed an edge intelligent gateway that can achieve multi-source heterogeneous data collection;Use 5G narrowband Internet of Things NB-Io T to directly upload data to the cloud,providing data support for subsequent data driven construction of neural network models;Using the Vision Transformer(ViT)algorithm in the field of computer vision as a benchmark model,the quality detection algorithm for pressed products made of rare earth magnetic materials is studied;In order to improve the quality detection effect of the network model on rare earth magnetic material molded products,based on the L2 regularization algorithm and the Dropout algorithm,this paper proposes an improved hybrid neural network algorithm(Hybrid algorithm),and conducts experimental designs on the publicly available data sets MNIST and CIFAR10 to verify the effectiveness of the algorithm proposed in this paper;In addition,according to the specific situation of the research object,the data preprocessing of the traditional ViT algorithm is improved,and the idea of image cutting is proposed;In order to make the obtained feature map have richer semantic information,a feature fusion module is introduced;Due to the insufficient number of training samples for rare earth magnetic materials,data enhancement methods such as Gaussian noise and salt and pepper noise are used to expand the number of training samples and improve the generalization ability of the network model;Finally,the multi-layer perceptron module of Vision Transformer is improved,introducing the Hybrid algorithm proposed in this article,and proposing the final improved ViT algorithm.Using the improved ViT algorithm and the traditional ViT algorithm,experiments were conducted on the rare earth magnetic material dataset,and the experimental results were compared and analyzed.The final experimental results showed that the detection accuracy of the rare earth magnetic material quality inspection algorithm proposed in this article was as high as 97.78%,which increased the detection accuracy by 53.12% compared to the traditional ViT algorithm,achieving improved control efficiency in molding production and improved product production efficiency,The goal of saving production and operating costs.It shows that the Hybrid algorithm and the improved ViT algorithm proposed in this paper are effective and feasible.