This study introduces a novel approach for identifying potential failure risks in missile manufacturing by leveraging Quality Inspection Management (QIM) data to address the challenges presented by a dataset comprising 666 variables and data imbalances. The utilization of the SMOTE for data augmentation and Lasso Regression for dimensionality reduction, followed by the application of a Random Forest model, results in a 99.40% accuracy rate in classifying missiles with a high likelihood of failure. Such measures enable the preemptive identification of missiles at a heightened risk of failure, thereby mitigating the risk of field failures and enhancing missile life. The integration of Lasso Regression and Random Forest is employed to pinpoint critical variables and test items that significantly impact failure, with a particular emphasis on variables related to performance and connection resistance. Moreover, the research highlights the potential for broadening the scope of data-driven decision-making within quality control systems, including the refinement of maintenance strategies and the adjustment of control limits for essential test items.

Recently, there has been an increasing attempt to replace defect detection inspections in the manufacturing industry using deep learning techniques. However, obtaining substantial high-quality labeled data to enhance the performance of deep learning models entails economic and temporal constraints. As a solution for this problem, semi-supervised learning, using a limited amount of labeled data, has been gaining traction. This study assesses the effectiveness of semi-supervised learning in the defect detection process of manufacturing using the MixMatch algorithm. The MixMatch algorithm incorporates three dominant paradigms in the semi-supervised field: Consistency regularization, Entropy minimization, and Generic regularization. The performance of semi-supervised learning based on the MixMatch algorithm was compared with that of supervised learning using defect image data from the metal casting process. For the experiments, the ratio of labeled data was adjusted to 5%, 10%, 25%, and 50% of the total data. At a labeled data ratio of 5%, semi-supervised learning achieved a classification accuracy of 90.19%, outperforming supervised learning by approximately 22%p. At a 10% ratio, it surpassed supervised learning by around 8%p, achieving a 92.89% accuracy. These results demonstrate that semi-supervised learning can achieve significant outcomes even with a very limited amount of labeled data, suggesting its invaluable application in real-world research and industrial settings where labeled data is limited.