Predicting aviation non-volatile particulate matter emissions at cruise via convolutional neural network

Area of Science:

• Atmospheric Science
• Environmental Engineering
• Aerospace Engineering

Background:

• Aviation is a significant source of high-altitude particulate pollution, impacting global radiative forcing.
• Existing models for aviation non-volatile particulate matter (nvPM) emissions require improvement for accuracy.

Purpose of the Study:

• To develop and validate a predictive model, APMEP-CNN, for aviation nvPM emissions.
• To assess the model's performance with sustainable aviation fuels (SAFs) and at different operational phases.

Main Methods:

• Utilized a convolutional neural network (CNN) technique for the APMEP-CNN model.
• Trained the model using aviation emission databanks and field study measurements.
• Incorporated fuel properties to account for the impact of SAFs on nvPM emissions.

Main Results:

• Achieved high accuracy in predicting nvPM emission index in mass (EI_m) and number (EI_n) at ground level (R² = 0.96).
• Demonstrated good predictive capability for cruise emissions (within ±36.4%) and SAFs (within ±33.0%).
• Validated the model's applicability for various engine types and fuel blends.

Conclusions:

• The APMEP-CNN model offers significant improvements over existing methods for predicting aviation nvPM emissions.
• This model can enhance global aviation emission inventories and inform assessments of aviation's impact on climate and health.
• Accurate nvPM emission predictions are crucial for understanding impacts on local air quality and global radiative forcing.