Machine Learning-Aided Design Optimisation(MLADO) in Vortex Shedding-Based Engineering Applications

Abstract

Computational design is a key part in most engineering applications, thanks to the possibility to create new designs in a safer, quicker and reliable environment. The recent developments in engineering are also guiding the classical design life cycle to a more sophisticated frameworks, such as the implementation of Machine Learning methods to support the design process. This work shows the potential of using the namely Machine Learning-Aided Design Optimisation framework to optimise vortex-shedding based applications, and it is applied as example to a vortex shedding aerodynamic-based design extendable to other applications. This framework consisted of using a predictive model to discard useless computations and speed up the efficient construction of surrogate models. The method is applied to the optimisation of a mechanical vortex shedding-based passive mixer achieving a successful design in terms of minimisation of pressure drop and maximisation of mixing efficiency.