Evaluation of Machine Learning Effectiveness in Classifying Cracks on Foundry Product Images

This article presents an analysis of the effectiveness of machine learning methods for classifying cracks on foundry product images. The study focuses on evaluating the model's ability to adequately determine the level of risk, which is critical for ensuring industrial safety and improving production efficiency. The model's performance was assessed using metrics such as accuracy, recall, and F1-score, as well as through the analysis of ROC curves and the area under the curves (AUC), providing a detailed examination of its performance under various conditions. The results of the study indicate a high potential for the model's integration into production lines; however, additional adjustments will be required to minimize false positives, including the application of fuzzy logic and the integration of non-destructive testing methods. The proposed methods and achieved results can be applied to the development of automated quality control systems in mechanical engineering and other industries to timely identify and prevent risks associated with casting quality.