Big Data Analytics for RFID-enabled Manufacturing: A Case Study on Production Efficiency and Quality Optimization

Abstract

This study presents a comprehensive big data analytics approach for processing and analyzing radio frequency identification (RFID) data in a smart manufacturing environment. Using a dataset containing 413,472 records of manufacturing operations, we extract actionable insights to improve production efficiency and product quality. Our methodology includes data preprocessing, cluster analysis, multi-dimensional worker efficiency evaluation, and machine learning-based quality prediction. The results reveal distinct clusters of manufacturing processes with varying levels of complexity, significant correlations between process sequences and equipment utilization, and a robust framework for identifying high-performing workers. Furthermore, our machine learning quality prediction system achieves an accuracy of 86.4%. These findings offer manufacturers data-driven strategies for optimizing worker assignments, process sequencing, and quality control in RFID-enabled environments. The results reveal five distinct processing clusters with varying complexity levels and quality characteristics. Comprehensive temporal analysis uncovered significant shift-based quality differences (ANOVA: F=8.55, p<0.001), with Night Shift achieving the highest quality (99.29%) and Day Shift the lowest (99.16%). A systematic weekly quality decline trend was identified (slope=-0.000833/week, R²=0.824, p<0.001), suggesting equipment degradation or material variation factors. Hourly analysis revealed peak quality at 05:00 (99.58%) and lowest at 07:00 (98.92%), corresponding to shift transition periods. Within-shift fatigue analysis showed minimal degradation, with only Night Shift exhibiting a slight decline (-0.15%).