Digital transformation (DX) is driving the implementation of smart factories in manufacturing, contributing to real-time analysis and optimization of process data. However, low data reliability can lead to quality degradation and increased costs, and problems such as missing values, outliers, and duplicate data can arise.
In this study, we developed an AI-based data cleansing and quality management framework, applied it to manufacturing process data, and aimed to improve data reliability through anomaly detection using machine learning. Through the analysis of Oil Gasket manufacturing data, we empirically evaluated the reduction of defect rates, improvement of production speed, and optimization of resources. We assessed the core elements of data quality, including accuracy, completeness, consistency, reliability, timeliness, and validity, and enhanced data integrity using AI models such as Random Forest, Autoencoder, KNN, and Multivariate Regression.
As a result of the research, it was confirmed that AI-based data quality management is effective in improving the productivity and quality of manufacturing, and the Data Quality Index (DQI) also improved by 6.5%.
Through the use of various AI models, it was confirmed that the Random Forest model, in particular, has excellent performance in classifying defective products. Future research will propose a smart factory operating model through real-time data processing and automation, and present more effective quality management methods by building an AI-based quality management system.