Facility Assessment of Indoor Air Quality Using Machine Learning

Facility Assessment of Indoor Air Quality Using Machine Learning

Primary researcher: Jared Wright

My thesis, titled ?Facility Assessment of Indoor Air Quality Using Machine Learning,? presents a methodology for identifying, monitoring, evaluating, and modeling indoor air pollutants in an electroplating manufacturing facility. The study focuses on particulate matter, total volatile organic compounds, and carbon dioxide. Analysis of long-term data against OSHA compliance standards reveals challenges in maintaining adequate particulate matter concentration. Machine learning techniques including support vector machines, neural networks, and Gaussian process regression were employed to model various areas of the building against ventilation, production, temperature, and humidity features. An exponential Gaussian process regression model emerged as the most suitable for representing the datasets based on validation/testing error and prediction standard deviation. Findings indicate the necessity of maintaining neutral humidity and supply-air flowrate during production for pollution reduction. Recommendations include redesigning parts of the ventilation system or installing a new exhaust system more suitable for the removal of particulate matter. Operational changes could also take place that reduce the amount of time pollutant sources are moved from their exhaust hoods.