Smart Buildings and loT research lab
Smart Buildings and loT research lab
- A new personalized comfort model based on wearable device data
We propose the development of a new comfort model based on the new wearable device technology. These smart wearables are equipped with a variety of high-precision sensors that enable the collection of extremely rich contextual information related to the wearer biometric data and his surroundings. Machine learning techniques, such Neural Net Work, are used to learn from these data a new personalized comfort model. In this case, the model initially takes personal comfort feedback from the end users either through a user’s direct report or through inference from human productivity measures. Then, the model gives the comfort level of each occupant associated with their human biometric measurement (e.g. heart rate and skin temperature). As a proof of concept, a user interface app to receive an occupant smart watch biometric data along with his direct feedback on comfort conditions is developed. Different machine learning algorithms and nonlinear fitting are then applied for training the best model to correlate the biometric data with occupant feedback.
- Low Refrigerant Algorithm Detection for Cooling Systems Relying on Trending and Data Analysis
In this work, a hybrid algorithm of an enhanced version of Mann-Kendall trending and data analysis is proposed to solve the limitations of current technology in detecting and diagnosing cooling system refrigerant faults in general and refrigerant leakage specifically. A data abstraction mechanism is applied over feed of temperatures and power measurement to calculate and store only the significant information for further analysis. Next, an enhanced version of Mann-Kendall trending is applied periodically over the stored data to calculate the trend strength (upward or downward) for each measurement. Finally, a harmonic mean is utilized to balance the trends contribution and evaluate the result against a threshold value for potential faults. Such an algorithm is expected to have an important positive impact, because it is designed to accurately detect low refrigerant at an early stage. This should help in the following ways: (a) to reduce the impact of refrigerant emissions on climate, and (b) to potentially reduce the U.S. energy use by more than 0.1–.02 quad per year. This algorithm is a robust first step towards leveraging the latest technology advancements, especially in computer science and mathematics, in order to vertically advance the field of cooling systems.
- Marine Transport Data Analysis and Failure Prediction (Funded by Emerson Climate Technology)
The objective of the pthis work is to analyze data log of multiple measurements and status conditions for multiple marine transports unites. Multiple techniques are planned for this investigation such as: (1) general data summary and analysis(2) machine learning methods to learn the system operation and measurements trends and use this learning to watch for any future deviation (3) observe the data during the failure events and compare with the most recent data before failure. Other objectives are to design algorithms for detecting low charge and air flow blockage.
- The PMV Thermal Comfort Model Sensitivity Analysis and the Use of Wearable Devices to Enhance Its Accuracy
This work studies the sensitivity of the Predicted Mean Vote (PMV) thermal comfort model relative to its environmental and personal parameters. PMV model equations, adapted in the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) Standard 55–Thermal Environmental Conditions for Human Occupancy, are used in this investigation to generate two-dimensional (2D) and three-dimensional (3D) comfort zone plots for different combinations of parameters. The metabolic rate for an occupant during normal life activities is recorded using a Fitbit® wearable device to demonstrate the PMV expected error range when personal parameters are ignored, and to determine the potential of using a wearable device to enhance PMV comfort model accuracy.
- Smart Connected Health Monitoring and Unintentional Injury Prevention Tool
The primary cause of death for children aged 1-4 and aged 5-9 years, respectively, is unintentional injury from drowning (31%, 15%), motor vehicle (25%, 49%), and fire (12%,11%). For children aged 10-14 the primary cause is traffic/transportation (83%) [CDC report, 2012]. Besides death, studies showed that marital disruption rate for bereaved parents is ~53% higher [Rogers et al., 2008]. The goal of the project is to reduce all childhood deaths by 15%-25% and reduce martial disruption rate. This projectaddresses the primary cause of death within the first years of life due to unintentional injury by giving the child a variation of a smart-band that tracks heart rate, oxygen saturation level, and body temperature, send these data continuously to the cloud, and alert parents if these measurements vary from the normal range based on some predetermined thresholds . This technology should also work outside the home and is targeted of kids above 1 year and could engage/address unintentional injury sources for children. For example, drowning and fire/heat could be detected and alerted in timely and effective manner. The idea of the project can be generalized to span other health monitoring and accident prevention systems by creating a basic sensor-based platform connected with a sufficient, secure, and scalable cloud implementation that will measure, store, candidate, fit and retain capable and user-specific medical predictive models, and present results including alerts for clear threat to health in a clear and informative manner to care providers including parents, medical responders, and transport personnel.