Research Fellows
Ismael Sànchez Torreguitart
Early Stage Researcher 13 at Von Karman Institute
Designing and optimizing a turbomachinery components is a complex iterative design process that can take significant time and effort. The use of relatively low-cost numerical shape optimisation methods has become more popular and have been widely used in turbomachinery applications. Within the different classes of optimization methods, the adjoint based methods have become increasingly popular amongst the gradient-based optimization methods for its efficiency to compute the gradients at a cost that is independent of the number of design variables. However, one major bottleneck herein is that adjoint methods mainly work on altering the shape to be optimized by deforming the CFD grid and as such loose the connection
to computer aided design (CAD), the industry adopted standard for the design of components.
This research work focuses on the integration of the CAD, the currently industrial accepted format for manufacturing, in a gradient based optimization framework. Defining the design variables in the CAD system requires computing the derivatives of the grid coordinates with respect to the CAD design variables. This was done by differentiating the CAD kernel and grid generator of the VKI’s in-house software CADO using the ADOL-C tool in forward and reverse modes of algorithmic differentiation.
In this way, it was possible to compute accurate gradients whilst avoiding the topology and label renaming issues that can arise with the finite difference approach. Using algorithmic differentiation for the derivative computation requires more memory than the finite differences approach, but it performs significantly faster specially for large number of design variables. The method is tested on the LS89 turbine cascade and large aerodynamic improvements in the entropy generation are achieved whilst keeping the exit flow angle fixed. The trailing edge and axial chord length, which are kept fixed as manufacturing constraints. Images of the baseline and optimal designs are shown below.
