Aircraft landing gear structures are exposed to complex loading in-service. Coupled with the geometry and joints used within landing gear structural assemblies, finite element models tend to be used to compute the loads acting on landing gear components during ground maneuvers. Concerning novel design approaches for complex structural assemblies, such as probabilistic assessment, optimization or ‘digital-twins’, the computational expense of using finite element models is prohibitive. Surrogate modeling methods have been proposed as a route to reducing the computational expense of assessing complex structural assemblies for static and fatigue design. This paper investigates the application of Response Surfaces, Radial Basis Functions, Gaussian Process Regression and Artificial Neural Networks as approaches to surrogate modeling for landing gear load models. Following the construction of the surrogate models within case studies representing a side stay and complex drag brace component, it was identified that Response Surface and Gaussian Process Regression surrogate models could be used to reduce the computational expense of a landing gear loads assessment from 20 seconds to less than a millisecond. As a result, surrogate modeling methods provide the required reduction in computational expense to support probabilistic design and optimization of complex structural assemblies.
|Conference||AIAA SciTech 2020 Forum|
|Period||6/01/20 → 10/01/20|