Thesis
Utilizing wearable sensors for estimating metabolic cost and evaluating individual protective equipment
University of Iowa
Master of Science (MS), University of Iowa
Spring 2026
DOI: 10.25820/etd.008462
Abstract
The US Navy utilizes different protective equipment depending on a given situation. This protective equipment, typically worn by damage control (i.e., emergency response) teams, potentially incurs a significant physiological burden on the human body. One measure of physiological burden is metabolic cost, defined as the energy expended to perform a specific task. Metabolic cost is most often estimated via indirect calorimetry, which measures the volumes of oxygen consumption and carbon dioxide production. However, indirect calorimetry requires participants to wear a mask, which is uncomfortable, usually restricts measurements to a laboratory, and produces noisy, low-frequency data. In some cases, such as when protective equipment already covers the face, a mask cannot be used. This study proposes using wearable sensors as an alternative approach. Wearable sensors provide many benefits including their relatively small size and rapid data collection in a wider range of environments. Existing approaches often rely on raw data and/or apply complex machine learning algorithms that may lack generalizability. To evaluate the physiological burden of protective ensembles, this research was conducted in two separate studies. The first study developed and validated a machine learning model to predict metabolic cost. The second study deployed this model to a novel set of participants who completed a series of tasks representative of naval damage control teams.
The first study developed and validated a machine learning model using an open-source dataset from Ingraham et al. (2019), which included 10 participants performing 6 exercises. For each trial, a participant completed the exercise for a 6-minute interval across different conditions (e.g., speed and incline). Input features were averaged over 15-second intervals and derived from 3 types of wearable sensors: heart rate (HR) monitor, inertial measurement units (IMUs), and electromyography (EMG) sensors. Following feature selection, the final set of features included mean HR (normalized by resting heart rate), a signed RMS numerical derivative of HR, a novel feature proposed for kinetic energy (derived from IMU data), summed EMG signals from the bilateral soleus and biceps femoris, and participant mass. Following a comparative analysis, the machine learning model selected was a support vector regressor with a linear kernel. The model was deemed to be successful when the 9-fold cross validation achieved an R^2 of 0.70 or greater and a normalized root mean squared error (NRMSE) of less than 0.15. The average R^2 across all participants was 0.87 and the average NRMSE was 0.076.
In the second study, a set of participants completed 6 tasks in a fixed sequence associated with the naval damage control teams: jogging, casualty drag, water wheel activity, fire extinguishing activity, stair climbing, and walking. Each task was performed in both a firefighting turnout suit and everyday athletic clothing (control), the order of which was randomized across participants. The study collected data using IMUs, EMG sensors, a heart rate monitor, and skin/core temperature sensors. Additionally, participant demographics, a perceived workload survey, and functional mobility screens were conducted. This data was analyzed using the machine learning algorithm from the first study to investigate how the protective ensemble influenced the physiological burden on the participant. The results showed a statistically significant increase (p < 0.05) in metabolic cost between the control and suit conditions for most tasks. The greatest increase was observed during the final activity (walking), suggesting that the estimated metabolic cost may be influenced as general fatigue accumulates. However, neither mobility nor perceived workload demonstrated a statistically significant correlation with changes in average instantaneous metabolic cost. These findings indicate that the observed increase in metabolic cost is likely driven primarily by the added weight and bulk associated with personal protective equipment (PPE).
Quantifying the physiological burden incurred during tasks performed by damage control teams is critical to ensuring their safety and optimal performance in high-risk environments. This work proposed a method of estimating metabolic cost using wearable sensors with the overall goal of reducing physiological burden of protective equipment on the user. By utilizing wearable sensors, data collection becomes more feasible across various environments. The findings of this work can provide insights to inform equipment design and training protocols.
Details
- Title: Subtitle
- Utilizing wearable sensors for estimating metabolic cost and evaluating individual protective equipment
- Creators
- Katherine M. Butler
- Contributors
- Rachel Vitali (Advisor)Shaoping Xiao (Committee Member)Kimberly Ingraham (Committee Member)
- Resource Type
- Thesis
- Degree Awarded
- Master of Science (MS), University of Iowa
- Degree in
- Mechanical Engineering
- Date degree season
- Spring 2026
- DOI
- 10.25820/etd.008462
- Publisher
- University of Iowa
- Number of pages
- xii, 62 pages
- Copyright
- Copyright 2026 Katherine M. Butler
- Language
- English
- Date submitted
- 04/14/2026
- Description illustrations
- Illustrations, graphs, charts, tables
- Description bibliographic
- Includes bibliographical references (pages 49-59).
- Public Abstract (ETD)
The US Navy utilizes different protective equipment depending on a given situation. This protective equipment, typically worn by damage control (i.e., emergency response) teams, limits mobility and incurs a significant physiological burden on the human body. One measure of physiological burden is metabolic cost – a measure of the energy expended to perform a task. Metabolic cost is often estimated via indirect calorimetry, which involves measuring oxygen consumption and carbon dioxide production. Despite its advantages and widespread use, indirect calorimetry requires participants to wear a mask, usually restricts measurements to a lab setting, and is often not compatible with protective equipment that covers the face. This work proposes using wearable sensors as an alternative, providing a more practical approach that can feasibly be integrated with protective ensembles and operational environments.
The overall goal of this thesis is to evaluate human performance while wearing protective ensembles during operationally-relevant tasks. The first study developed and validated a wearable sensor-based model for estimating metabolic cost, which was subsequently used in the second study to quantify the physiological burden of protective equipment during operationally relevant tasks. The model was developed using an open-source dataset and achieved an average R2 of 0.87 and average NRMSE of 0.076 with a 9-fold cross-validation. Given the success of that model, a second study was conducted in which participants were outfitted with a wearable sensor array and completed 6 tasks in both a firefighting turnout suit and everyday athletic clothing. The wearable sensor data were used as inputs to the aforementioned model to determine if the protective ensemble significantly changed the physiological burden on the participant. The results showed a significant increase in metabolic cost was incurred for most tasks when the protective suit was worn compared to the athletic clothing. While participants experienced only a small decrease in mobility, their perceived workload increased. However, these changes showed no significant correlation with the increase in metabolic cost, suggesting that the increased burden is mainly caused by the additional weight and bulk of the protective equipment. The findings of this thesis can provide insights to inform equipment design and training protocols.
- Academic Unit
- Mechanical Engineering
- Record Identifier
- 9985177376702771
Metrics
1 Record Views