07-10, 10:45–11:15 (US/Pacific), Room 315
Aviation accounts for 2% of global greenhouse gas emissions, and reliance on liquid petroleum-based fuels makes this sector challenging to decarbonize. We seek to accelerate the development of sustainable aviation fuels using an early-stage design tool with a data-driven approach. We developed our strategy using the Python-based optimization packages BoTorch and Ax, and also rely on Pandas. We will discuss how to down-select from many possible fuel components to a specified number of chemical species and identify which combinations are most promising for a novel sustainable aviation fuel. We will also present its integration in our open-source web tool supporting biofuel research.
Aviation represents 2% of global greenhouse gas emissions. Sustainable aviation fuels are necessary to decarbonize the industry and help achieve climate goals. However, certifying candidate fuels is a resource- and time-intensive process. Fuels with properties, such as boiling point or kinematic viscosity, that fall outside of specifications are either rejected part-way through certification or incur further costs to correct. This demonstrates a need for rapid and inexpensive computational models to assist in early-stage design of sustainable aviation fuels.
In this presentation, we will discuss how we use Python to identify fuel component combinations that achieve desired property target values. We will focus on challenges related to complex optimization problems and massive search spaces as well as scientific Python tools that assisted us in developing this strategy.
The primary challenge of designing sustainable aviation fuels is the size and complexity of the search space; there is one dimension for each possible fuel component which makes sampling through the whole search space very time- and computationally intensive. To reduce the search space and accelerate the process, we sample discretized volume fractions of fuel components that add up to 100% with the help of itertools. Next, we estimate the expected values for two properties of each permutation and use Pandas to both organize the permutations and rank them. Through this process we identify the best-performing, preliminary fuel component combinations which we then pass to our lower-level optimization routine.
The lower-level optimization finds the best volume fraction composition of a fuel with respect to provided properties of interest. We employ the open-source, multi-objective Bayesian optimization packages BoTorch and Ax for our method. These packages emulate the underlying functions of the best-ranking fuel combination via a surrogate model. The optimization progresses when an acquisition function identifies where in the search space the surrogate will iteratively update itself. BoTorch and Ax then use the surrogate model to return the volume fraction composition that best optimizes the properties of interest.
To help accelerate sustainable aviation fuel research, we have integrated this strategy into our web tool supporting biofuel research for open-source access. We also aim for our work to further the exploration of complex optimization methods and their open-source tools for other relevant applications beyond sustainable aviation fuels.
Ali Martz is a Graduate Research Assistant pursuing a M.S. in Mechanical Engineering at Oregon State University. Her research focuses on the optimization of sustainable aviation fuel design.
