An AI-powered data processing framework for RFID-captured manufacturing datasets

With the rapid development of artificial intelligence (AI), the demand for data has been surging. More attention has been paid to data in their daily processes, such as the production processes. Deployed in manufacturing sites to control and monitor processes, the Internet of Things (IoT) technology specifically radio frequency identification (RFID) in industrial settings has shown its potential as a data collection approach. However, the data collected by the RFID suffers from several challenges such as duplication, missing data, etc. Therefore, this paper focuses on the development of a data processing framework for addressing the challenges. The framework will process real RFID-captured production data from an IoT-enabled manufacturing shop floor with three functionalities: data pre-processing, outlier detection, and big data analytics. For anomaly detection, this framework deals with the passing rate of different production processes with a detection model, which can be used to flag abnormal production cases to facilitate the quality control process. The flagged abnormal production cases will be generalized during big data analytics to investigate the reason behind underperformance.