Thermal design optimisation in turbomachinery elements
Design optimization techniques are a new frontier in engineering planning. Usually designers use “cut and try” methods in developing their projects, taking advantage of their Company’s Databases and internal know-how. Optimization algorithms can additionally support these systems and give a relevant improvement to the development time and design quality. In its R&D department, AVIO Group is implementing a number of projects in order to test the capabilities of modeFrontier, when coupled with its own internally developed 1-D CFD software (BLADECOOL) applied to turbo machinery problems. The 1-D code was developed in collaboration with several Universities and was validated in several heat transfer problems by AVIO Group in the past decade. The software computes the air mass flow and the heat transfer coefficient on the internal surface of the blade. Due to the high non-linearity of the system, small changes in the diameters or in the jets distribution produce relevant changes in the cooling performances. Therefore, the company used slight perturbations of the standard design in order to improve the cooling efficiency.
In this paper we illustrate some results obtained applying these methodologies to the aero-thermal field, and, in particular, to the turbine blade cooling design. The cooling systems of turbine blades use a percentage of the compressed air that cannot so be used in the main engine thermodynamic cycle. The reduction of this percentage, without decreasing the cooling effectiveness is one of the most important tasks in blade design. In fact, this reduction would grow up the whole engine performances. We show the results obtained applying a Genetic Algorithm to an impingement cooling system acting on a simple test geometry. The MOGA algorithm found a set of designs which at the same time reduce the mass-flow and increase both the heat transfer coefficient and its uniformity on the blade surface. After the multi-objective phase the AVIO designers decided to turn the HTC objectives into constraints and to look for solutions which reduce the compressor spilled air mass-flow. The single-objective phase was approached both with the MOGA and with the SIMPLEX algorithm obtaining the same results. The final single-objective phase found solutions which keep the same HTC average value of the base design and which considerably reduce both the air mass-flow and the HTC standard deviation.