Predicting Conceptual Aircraft Design Parameters Using Gaussian Process Regressions on Historical Data

This paper presents a novel methodology for predicting key aircraft design parameters using Gaussian process regressions (GPRs) applied to a comprehensive, open-source database of over 400 aircraft and 200 engines. The database, made freely available through the Future Aircraft Sizing Tool (FAST), enables aircraft designers to apply detailed historical data in early-stage conceptual design and provides full visibility into the data underlying the regressions—unlike traditional regression models, where the training data and model fit are often not disclosed. The nonparametric GPR models developed in this work allow for flexible input configurations, improving the accuracy of predicting critical parameters, such as operating empty weight and uninstalled engine weight, compared to established regressions from the literature. By incorporating a broader range of inputs, these models reduce prediction errors and provide tighter error distributions, leading to more reliable estimates during early design phases. This paper outlines a methodology for adapting conventional aircraft data to explore hybrid-electric and fully electric aircraft designs, ensuring that historical data can be leveraged effectively for novel propulsion systems. Additionally, the flexibility of the GPR framework allows users to create their own regressions and update predictions as new data becomes available, making it a useful tool for researchers working with their own datasets.