Predicting Respiratory Health Symptom Occurrence in Office Building Environments
Open Access
- Author:
- Vukovic, Vladimir
- Graduate Program:
- Architectural Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 10, 2009
- Committee Members:
- Dr Jelena Srebric, Dissertation Advisor/Co-Advisor
Jelena Srebric, Committee Chair/Co-Chair
Stanley Allan Mumma, Committee Member
Bohumil Kasal, Committee Member
Leonard J Peltier, Committee Member
Zhengmin Qian, Committee Member - Keywords:
- artificial neural networks
perceived environmental quality
BASE benchmark database
indoor air quality
respiratory health symptoms
sick building syndrome - Abstract:
- Providing healthy indoor environments for building occupants is important to ensure people's wellbeing, satisfaction and productivity. Furthermore, healthy indoor environments can reduce health care costs and positively affect the economy. Currently used techniques to establish relations between indoor environmental parameters and occurrence of health related symptoms among occupants cannot deal with real time concurrent changes of multiple simultaneously monitored parameters, nor provide predictions of real building impacts on occupants' health. The dissertation is advocating usage of innovative data analyses tools for interpretation of building sensor data and occupants' perceptions of indoor environments to predict respiratory health symptom occurrence in buildings. Such data analyses establish relations between indoor air quality perceptions and health. The research goal was development of an artificial neural network (NN) based computational methodology for fast prediction of respiratory health related symptoms among office building occupants. For the purpose of NN training, Environmental Protection Agency's (EPA) Building Assessment Survey and Evaluation (BASE) study provided measurements of indoor building parameters and occupant survey data within 100 office buildings in the U.S.A. Trained networks had an output indicating occupants' health symptoms based on measurements and occupants' perceptions of indoor environments. The method was tested and experimentally validated using on-site measurements and occupants' survey for a LEED certified "green" building environment. Additional multivariate statistical regression of the BASE data was used for the purpose of comparing results to presently available data analyses tools. The results showed NN methodology can be applied to predict a number of respiratory health symptoms among office building occupants. High significance of occupants' perceptions of indoor environments was confirmed by multivariate statistical analyses. Experimental study in a green building revealed better indoor environmental quality, healthier indoor conditions and higher occupants' satisfaction compared to an average BASE office building. The developed methodology could be incorporated in the future design procedures to specify optimal combination of indoor environmental parameters and prevent possible adverse impacts on occupants' respiratory health. Last, but not the least, the results encourage building researchers and scientific community to initiate applications of NNs, as innovative and powerful data interpretation tools for indoor environments.