Distributed computing and cutback trailing edges

It is the received wisdom within industrial design departments that large eddy simulation type calculations are, due to their expense, best relegated to performing a few high fidelity calculations which can then be used to inform and calibrate lower order methods. It is shown here that there is a class of flow for which this is an inappropriate method of usage. A member of this category of flows is the cutback trailing edge flow - effectively, a wall jet with added complexity. These have been shown to be highly intractable to steady RANS-type solutions, but relatively inexpensive and easy to simulate with LES-type methods. Here, large eddy simulation on coarse grids is found to give comparable accuracy to, at around 10% the cost of, fully resolved LES simulations. To demonstrate the potential power of careful application selection, an optimisation exercise on cutback trailing edge designs was carried out using VLES. In order to take advantage of the enormous parallel capacity of moden supercomputers and distributed computing networks, a perfectly parallel evolutionary algorithm was implemented to direct the optimisation. As a demonstration, a fairly crude single target optimisation was carried out, which aimed to optimise the film cooling effectiveness integrated over the cutback surface. This was to be achieved by varying the layout of the internal cooling structures within the coolant cavity. Although the turbulator layouts were allowed to vary significantly, there were some functional restrictions placed upon possible designs. Six hundred LES-type simulations were then carried out, as the evolution progressed over 12 generations. An optimised design was then extracted from the final generation, which showed significant improvements in film cooling over the performance of the planforms used in the original experiment. The influence of various design parameters on the film cooling effectiveness could then also be indirectly explored by extracting clouds of data from the simulation results. It is likely that more advanced optimisation heuristics could have used this data to accelerate convergence to optimised designs.

[1]  P. Martini,et al.  Experimental and numerical investigation of trailing edge film cooling downstream of a slot with internal rib arrays , 2003 .

[2]  J. Peter,et al.  Comparison of surrogate models for turbomachinery design , 2008 .

[3]  S. A. Harvey,et al.  The design optimisation of turbomachinery blade rows , 2003 .

[4]  Xin Yuan,et al.  Towards robust unstructured turbomachinery large eddy simulation , 2015 .

[5]  Paul G. Tucker,et al.  Linear and nonlinear model large-eddy simulations of a plane jet , 2008 .

[6]  Antony Jameson,et al.  Formulation of Kinetic Energy Preserving Conservative Schemes for Gas Dynamics and Direct Numerical Simulation of One-Dimensional Viscous Compressible Flow in a Shock Tube Using Entropy and Kinetic Energy Preserving Schemes , 2008, J. Sci. Comput..

[7]  T. Simpson,et al.  Comparative studies of metamodelling techniques under multiple modelling criteria , 2001 .

[8]  J. Smagorinsky,et al.  GENERAL CIRCULATION EXPERIMENTS WITH THE PRIMITIVE EQUATIONS , 1963 .

[9]  Andy J. Keane,et al.  Comparison of Several Optimization Strategies for Robust Turbine Blade Design , 2009 .

[10]  Leigh Lapworth,et al.  PADRAM: Parametric Design and Rapid Meshing System for Turbomachinery Optimisation , 2003 .

[11]  Meng-Sing Liou,et al.  Transonic Axial-Flow Blade Optimization: Evolutionary Algorithms/Three-Dimensional Navier-Stokes Solver , 2004 .

[12]  Timoleon Kipouros,et al.  Multi-objective Optimisation of Turbomachinery Blades Using Tabu Search , 2005, EMO.