A low-cost IoT architecture with RF communication and data analytics for real-time productivity monitoring in garment...

Productivity control in garment sewing lines is a critical factor in ensuring on-time delivery and competitiveness in the apparel industry. In many developing countries, however, line-level monitoring remains manual, often performed at the end of the day, which leads to errors, a lack of real-time visibility, and a limited ability to intervene during operations. To address these challenges, this study presents the design and implementation of a low-cost IoT architecture for real-time productivity monitoring in sewing lines. The proposed system integrates hardware modules at the station and line levels, a radio frequency (RF)-based communication protocol, firmware for data acquisition, a local desktop interface, and a web-based management platform. Beyond monitoring, the study leverages machine learning to enhance decision-making. A random forest regression model is employed in a recursive forecasting framework to predict output in the coming hours and end-of-day production based on partial-day outputs and assigned targets. This approach enables supervisors to anticipate shortfalls early in the shift and take corrective actions such as line balancing or adjusting targets. The system was deployed in a garment factory for experimental validation. Results demonstrate that the RF communication achieved stable performance with low latency and over 98% transmission success after retries. The random forest model achieved high predictive accuracy, with R ² = 0.997 on the training data and R ² = 0.963 on the test data. Benchmarking against kNN, XGBoost, linear regression, and long short-term memory (LSTM) using MAE, RMSE, and training time shows that random forest achieves higher accuracy than kNN, XGBoost, and linear regression while training substantially faster. However, slightly less accurate than LSTM, it trains much more quickly, representing the best accuracy-efficiency trade-off for real-time deployment. Experiments also show that the prediction results from random forest help to balance the lines effectively and adjust the goals of the lines, confirming that the proposed system enables both reliable monitoring and proactive management.