CWMformer : a novel framework for long-term fatigue conditions measurement and prediction of construction workers with a wearable optical fiber sensor system

Abstract

Construction workers face numerous dangers and high-intensity environments at work. Monitoring their vital signs and fatigue is crucial for ensuring their health and safety, preventing accidents, and improving work efficiency. Optical fiber sensors are widely used due to their small size, lightweight, and strong resistance to electronic interference. These sensors are particularly useful for detecting and monitoring the vital signs of construction workers. Heart rate variability (HRV) is the variation in consecutive heartbeat intervals and is a marker of autonomic nervous system activity, influenced by various factors. Fatigue significantly impacts HRV and overall cardiovascular health, making HRV a useful tool for assessing fatigue conditions. Monitoring HRV helps identify fatigue conditions. However, existing optical fiber sensors have drawbacks such as signal loss, distortion during long-term transmission, and high manufacturing and maintenance costs. They may also produce missing or anomalous data due to sensor failures and other unforeseen factors, making long-term fatigue measurement challenging, especially when direct contact with the skin causes discomfort. To address these issues, we propose a novel optical fiber sensor system based on a fiber interferometer integrated with smart clothing. In addition, we propose a robust deep learning framework called CWMformer, which processes and analyzes raw vital signs of construction workers, predicts long-term fatigue conditions, and matches them with HRV more accurately. Our experiments show an average mean absolute percentage error (MAPE) of 2.126 and a p-value of 0.0213, indicating feasibility and effectiveness. This study presents a novel and practical application for smart healthcare monitoring of construction workers.