Machine Learning for the Development of Data Driven Turbulence Closures in Coolant Systems

Turbine Entry temperature is increasing and the coolant systems are becoming more and more complex. With the advent of Additive Manufacturing, it is possible to use geometrical configurations that were unthinkable up to few years ago. Unfortunately design methods used for the coolant systems still relies on RANS modelling, being two orders of magnitude faster than DES simulations. There is a clear need for accurate RANS simulations able to mimic the accuracy of DES models without the computational cost.

This work shows the application of Gene Programming for the enhancement of turbulence closure model in a CFD solver for a complex geometry designed for additive manufacturing.

One of the challenges in internal coolant design optimization is the heat transfer accuracy of the RANS formulation. DES is inherently more accurate but it is still not used for design. This work shows a Data Driven approach to develop a turbulence closures for internal ribbed duct. 

Different approaches are compared and the results of the improved model are shown. In particular it is shown how the inlet region is particularly critical for the training, due to the complexity of the geometry.

The work shows the potential of using Data Driven models for accurate heat transfer predictions even in non conventional configurations.