Prioritizing odorous VOCs emitted from air filters using machine learning models

Area of Science:

• Environmental Chemistry
• Analytical Chemistry
• Computational Chemistry

Background:

• Ozone reactions with air filter pollutants generate odorous volatile organic compounds (VOCs).
• Evolving pollutants and intensified ozone reactions can create new compounds, worsening odor problems.

Purpose of the Study:

• To develop and train a machine learning (ML) framework for predicting odor thresholds of VOCs emitted from air filters.
• To identify key VOCs contributing to odor on air filters for targeted monitoring and control.

Main Methods:

• Trained ML models (Random Forest, Bagging Regression, Gradient Boosting) using datasets of 874 VOCs and 240 properties.
• Analyzed VOCs from two types of used air filters using GC-MS and HPLC under varying ozone levels.
• Calculated odor thresholds and odor activity values for identified VOCs.

Main Results:

• Ozone significantly increased VOC emissions from filters (1.1-1.6x more detected VOCs, 2.1-2.9x higher total VOC concentrations).
• The Random Forest model demonstrated superior performance with R² = 0.786 and RMSE = 0.657.
• Aldehydes were identified as the primary contributors to odor based on odor activity values.

Conclusions:

• Machine learning provides an effective framework for predicting VOC odor thresholds from air filters.
• Ozone exposure exacerbates odor issues by increasing VOC emissions from air filters.
• Targeted monitoring and control strategies can be developed based on identified odor-causing VOCs like aldehydes.