| Summary: | This thesis stands for a 2D axisymmetric study of convection heat transfer, within cooler ducts of dry type transformers, by means of CFD simulations (ANSYS Fluent). The purpose was the definition of descriptive functions for the mean heat transfer and friction coefficients, regarding free and forced convection, within a vertical cylindrical annular duct, under isoflux conditions (uniform wall heat flux). Dry-transformer type uses air as cooling system by natural convection and an AC system is assumed at the base of the transformer for forced convection. All system conditions and fluid properties were constructed and defined at Fluent program. Several pre-tests were performed, mainly in order to properly select the studied mesh and fluid-dynamic model. The parametric study was assembled accounting for 45 geometry’s designs, 8 values for wall heat flux, 4 velocities for forced convection simulations, while for natural convection a linear correlation heat flux-velocity was assumed. This corresponds to 1800 simulations. After data analysis and treatment, a non-linear regression was performed in MATLAB, aiming for the descriptive functions of convection heat transfer and friction coefficients, obtained thru CFD. A successful fitting was obtained through artificial neural networks (ANN), providing predictions for convective coefficient of relative errors inferior to 12% for approximately 80% of cases and inferior to 15% for approximately 70% of cases for friction coefficient. It was concluded an overcome accuracy of the ANN’s predictions shown, compared to the most fitted literature’s model considered, for convective heat transfer coefficient.
|
|---|