Ali Martz et al. - Python for early-stage design of sustainable aviation fuels | SciPy 2024

Sustainable aviation fuels (SAFs) face major challenges in implementation, including high costs, complex certification processes, and insufficient production volumes to meet market demand

Aviation accounts for 2-3% of global CO2 emissions and is difficult to decarbonize due to the high energy density requirements of aircraft fuel

The fuel certification process requires 3-5 years, thousands of liters of fuel, and millions of dollars, with many candidates failing due to incorrect property values

The team developed a sequential optimization methodology using Python packages (Botorch, Axe) to:

• Reduce search space to identify best-performing components
• Optimize blend ratios using multi-objective Bayesian optimization
• Predict fuel properties using machine learning models

The optimization tool considers multiple critical properties including:

• Freezing point
• Flash point
• Viscosity
• Density
• Boiling point

Key features of the tool include:

• Flexibility to require specific components
• Ability to set blend ratio limits
• Modular design allowing different property predictors
• Integration with lifecycle and techno-economic analysis tools

The methodology successfully identified viable fuel blends meeting target property values while maintaining required Jet A fuel compatibility

The approach aims to accelerate SAF development by reducing resource requirements and identifying promising candidates earlier in the development process

Open-source Python tools enable complex multi-parameter optimization while handling the inherent variability of biomass feedstocks

The solution connects early-stage design to production pathways, helping bridge research and commercial implementation