Numerical prediction of flow, thermal and stress fields in gas turbine combustor components
In this work an integrated set of numerical methods is developed for the analysis of gas turbine combustors, which can predict the flow, temperature and stress fields in modern geometrically complex combustor walls. A key problem for accurate flow and temperature field prediction is the wide range of geometric length scales within modern combustor components. These components typically contain multiple small-scale cooling features such as pedestals and effusion cooling holes, which cannot be resolved by a computational mesh without incurring huge penalties in terms of computer processor and memory requirements. In this work a sub-grid-scale model is developed, which accounts for the effects of small-scale features such as pedestals without resolving them in the computational mesh. Validation of this model using experimental results from the literature shows that the pressure drop, turbulence generation and heat transfer effects of pedestal arrays can be successfully modelled using this approach. Another difficulty in the analysis of combustors is coupling the interdependent temperature field predictions in the fluid and solid regions. This has led to a unified approach to conjugate heat transfer prediction being adopted in this work, whereby a structured finite volume solver is used to predict temperature fields throughout fluid and solid domains. A new conjugate heat transfer discretisation scheme is developed, which can cope with the demanding combination of strong temperature gradient discontinuities and highly skewed grids. Several test cases are presented which demonstrate the accuracy of this new scheme, as well as demonstrating the inadequacy of conventional treatment of the diffusive fluxes for the solution of conjugate problems. The assembled numerical methods are used to predict the flow, thermal and stress fields in a geometrically complex combustor heatshield/backplate assembly, typical of that found in modern engines. This calculation shows that a viable route to computational life prediction has been established.