Synthesis of the Artificial Intelligence and Model-Based and Statistical Algorithms in the Classification of the Metal Surface Defects

Abstract

Steel has played an indispensable role in numerous industries, particularly in architecture, aerospace, and the automotive sector, and has been one of the most crucial components in manufacturing. The possibility of defects in the steelmaking process has had a substantial impact on the quality and service life of the final product. With the objective of ensuring a timely response in steel production, this paper has presented a model for the classification, detection of defect regions, and visualization of spatial defects. The model has been founded on the synthesis of convolutional neural network, snake algorithms, and algorithms for generating spatial defects based on images. The convolutional neural network has been trained using images from the NEU Surface Defect database, and model evaluation has been carried out on previously unseen samples that have not been included in the training data. The convolutional neural network has achieved an overall accuracy of 88.4% with unseen samples from the NEU Surface Defect database, with predictive abilities ranging from 72.7% to 97.7%. Following the classification, a spatial representation of the damage has been generated, and defect segmentation on the material has been executed. The application of this model in modern industry has the potential to significantly enhance the performance and quality of high-risk manufacturing processes, mitigate unnecessary losses, and enable informed decision-making about future steps in a more insightful manner.