The skin of modern aircraft is composed of multiple layers of protective and decorative paints which must withstand high environmental loads during flight. Due to various mechanical and structural properties involved, optimizing paint systems for aircraft is a challenging task.

In this work, we present an integrated, multi-scale machine learning approach to optimize aircraft paint systems. On an atomistic scale, we developed a machine learning model for predicting relevant mechanical properties of epoxy and polyurethanes—the base components of paints for aircraft. On a microscale, we developed a machine learning model for predicting peak mechanical stresses within a multi-layer paint system. We use both models in conjunction with a genetic algorithm to optimize layer thicknesses, mechanical properties, and the structure of the base polymers of the paint system. All models are deployed as an interactive web tool.

The presented approach enables the inverse design of aircraft paint systems, i.e. derive optimal paint system configurations for their specific applications and operating conditions, from the microscopic to the macroscopic scale in real-time.