File(s) under embargo
until file(s) become available
ARTIFICIAL INTELLIGENCE POWERED PHYSICAL FATIGUE ANALYSIS OF TRANSPORTATION WORKFORCES
The transportation industry has accounted for various work-related musculoskeletal disorders (WMSDs). A critical contributing factor to these WMSDs is physical fatigue occurring during activities that require the workers to adopt awkward postures and to exert excessive forces repetitively. While extensive previous research has been conducted in the construction sector to study injured workers, common WMSDs, related risky activities, ergonomic interventions, and physical fatigue detection methods to ensure the health of workers, there was a notable gap in understanding how physical fatigue and WMSDs can be prevented in the transportation industry. Due to the difference of duration, intensity, frequency, and procedure of activities between the transportation and the construction industries, findings from the construction industry cannot be directly applied into the transportation industry. Therefore, it was necessary to explore how the risk of developing physical fatigue and WMSDs can be mitigated, specifically in the transportation industry.
Preventing the development of physical fatigue and WMSDs in the transportation industry requires a comprehensive study including identification of worker types who were injured the most, exploration of the most common WMSDs and the activities causing the most injuries, proposal and evaluation of ergonomic interventions, and automated detection of physical fatigue. Specifically, it was found that transportation maintenance workers were injured the most based on the historical injury data of transportation workers from the Indiana Department of Transportation. Second, it was found that lower back injuries were rated as the most common WMSDs and lifting activities (e.g., lifting bags of materials and sign stands) were rated as the activity causing most WMSDs, followed by pushing/pulling activities (e.g., shoveling and pulling a dead deer), based on an online survey of transportation maintenance workers. Two subsequent on-site observations were conducted to gain a detailed understanding of how exactly the riskiest activity (i.e., lifting bags of dry concrete mix) was exactly performed and to suggest ergonomic interventions. Third, two ergonomic interventions were proposed based on unsafe postures that workers used and the excessive weights of loads that workers lifted. Specifically, given that workers needed to stoop (bending their backs) to lift bags of dry concrete mix placed at a lower level in real practice, a back exoskeleton could potentially support the workers in maintaining a healthier posture and in providing force to the worker’s back muscles when they bend their backs beyond a safety range, thus lowering the chance of excessive back muscle exertion, and then reducing the risk of developing physical fatigue and WMSDs. Additionally, establishing and adhering to a recommended lifting weight limit could alleviate the excessive exertion of workers muscles, thereby contributing to a decrease in physical fatigue and WMSDs prevalence.
To test the effectiveness of the two proposed ergonomic interventions, field experiments were performed. The experiment involved 29 transportation maintenance workers who lifted bags of dry concrete mix of three different weights including 31.5-pound bags (a weight calculated based on the Recommended Weight Limit equation considering actual vertical distance, horizontal distance, angle of asymmetry, etc.), 50-pound bags (a weight calculated based on the Recommended Weight Limit equation under the ideal condition, and it is also the weights sometimes used onsite), and 80-pound bags (the most common weights used on site) with and without wearing a back exoskeleton. The workers’ physical fatigue was measured using the surface electromyography (EMG) sensors, the electrodermal activity (EDA) sensors, and the heart rate (HR) sensors, while their perceived exertion of muscles, local perceived pressures, and acceptance levels were measured through subjective scales. The results from field experiments revealed several key findings: (1) There was a significant decrease in lumbar erector spinae (LES) muscle activities when wearing the back exoskeleton, especially during lifting 80-pound bags (from 61.7% to 50.8% of EMG); (2) when the weight of the bags was reduced to 31.5 pounds, there was a significant decrease in LES muscle activities (from 61.7% to 16.8% of EMG), in comparison to the activities measured while lifting 80-pound bags; (3) among all muscle groups, the LES activities were the most significant during lifting; (4) using a back exoskeleton helped to lower the perceived muscle exertion during the lifting of bags of varying weights; and (5) a majority of participants deemed the back exoskeleton as an acceptable form of assistance.
In addition to the adoption of ergonomic interventions, the detection of physical fatigue at an early stage can also be used to alert workers to rest or adopt correct postures, thus effectively preventing the development of WMSDs. Based on the field experiments, the approach to physical fatigue detection was explored by using different machine learning algorithms. Participants' physical fatigue levels were measured through EMG, EDA, HR, and motion (collected by inertial measurement units) data. A 10-fold cross validation revealed that the Artificial Neural Network (ANN) algorithm achieved the highest accuracy (87.84%) in identifying and predicting three physical fatigue levels (i.e., the low physical fatigue level, the medium physical fatigue level, and the high physical fatigue level) among transportation maintenance workers.
The findings from this research can aid transportation professionals in better understanding the risks in their workplace and provide better solutions. For example, transportation professionals should pay more attention to the identified types of workers, areas of injury on the body, and activities causing most WMSDs. This research also provided evidence that the back exoskeletons and the 31.5-pound bags can effectively reduce the risk of having physical fatigue. Therefore, transportation professionals could consider implementing the back exoskeletons or replacing current 80-pound bags with 31.5-pound bags to assist workers to perform activities in a safer way. In addition, this research provided a tool for automated physical fatigue detection which can be potentially utilized to alert workers of physical fatigue at an early stage. Theoretically, this research provided insights into the possibility of employing physiological measurements as effective tools for evaluating ergonomic interventions in real-world contexts. Furthermore, this research delved into the use of the ANN algorithm, integrating with EMG, HR, and motion data, and proved their accuracy in detecting physical fatigue among transportation maintenance workers.
This research also has several limitations and needs future work. For example, the benchmark used in this study to label three physical fatigue levels (i.e., the low physical fatigue level, the medium physical fatigue level, and the high physical fatigue level) was the Borg 6-20 scale, which is a subjective measure and may be susceptible to recall bias. Monitoring of lactic acid levels, as a gold standard for measurement of physical fatigue, could be employed in the future to offer a more accurate measurement of physical fatigue levels for the physical fatigue detection. However, measuring lactic acid comes with its ethical challenges, due to the invasive nature of the test. Therefore, finding a balance between obtaining accurate and objective measures and ensuring the well-being and consent of the participants can be a critical component of future research.
- Doctor of Philosophy
- West Lafayette