I-Corps: Predicting and Preventing Mold Growth and Unforeseen HVAC Equipment Failures with an Intelligent Monitoring and Alerting System

Vanderbilt University, Nashville TN

Investigators

Abstract

This I-Corps project will increase a customer’s ability to identify and mitigate mold and HVAC issues before they happen, allowing them to better allocate their resources and ultimately save money and maintain a healthy atmosphere and reduce waste. The core technology is centered around a holistic approach to indoor air quality (IAQ) and HVAC equipment maintenance. More than 75% of IAQ problems are HVAC related, and poor IAQ results in reduced productivity of approximately 14.5 million missed workdays a year costing the US more than $168 B annually. Widespread adoption of this technology would help companies increase their productivity due to better indoor air quality. Several studies have concluded that poor indoor air quality results in reduced performance. Some of the studies have calculated the total costs associated with mold-related asthma and sinus issues alone to be up to $32 B in medical bills and absenteeism. The approach focuses on HVAC equipment and enabling preventive maintenance, which may save the manufacturing industry over $200 B annually. This I-Corps project is focused on developing an intelligent monitoring and predictive system that is capable of 1) predicting future mold growth, 2) making predictions of future HVAC system health, and 3) generating alerts, work orders, and maintenance recommendations to the end user. Such a service does not currently exist within traditional building automation systems, which only monitor the mechanical systems with a focus on energy optimization, not the environment with a focus on occupant comfort. Facility managers continue to take a reactive approach to HVAC system maintenance and mold remediation, rather than a proactive, preventative approach. The goal is to evaluate several approaches to quantify the state of health of HVAC systems with limited data under the assumption that access to proprietary data is not possible. Time series analysis methods, such as vectorized auto regression, provide accurate anomaly detection results under a short prediction horizon, but are not suitable for fault isolation. Using mold growth and rate charts from the literature, a machine learning algorithm has been developed to accurately classify the risk level of a given temperature and humidity reading, but further work is needed to account for a cumulative exposure rate or measurement fluctuations. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.