These simplifications lead to separate analyses being performed on the external and internal flow paths, blade metal and surface heat transfer. It creates the need to iterate many times between the separate analyses, updating boundary conditions up to a point where the interdependency between them converges. Often, simplifications to geometries are so extreme that experimental correlations are needed so as to force validity of simplified modeling. This article describes an optimized workflow that allows engineers to work with actual geometries and to combine flow, thermal and ultimately stress analysis into a single model. The framework is exemplified here on a shrouded-blade production geometry, that originates from the Tornado engine, a 5-7MW powergeneration unit from Siemens (formally Ruston Gas Turbines), which would now typically be undergoing overhaul. The blade contains a curvilinear internal cooling path exiting at the shroud. Details such as dust holes need to be represented faithfully. This engine is maintained by Wood Group light industrial turbines.

Meshing Strategy
A combination of surface wrapping and surface re-meshing techniques were used to repair flaws in CAD surface representation. At the same time, feature detail such as film cooling holes and high curvature are retained. The CAD was exported in IGES format and read into STAR-CCM+ for meshing and analysis. In this example, over 600,000 errors in the IGES tessellation were found, ranging from unclosed surfaces to excessively large aspect ratios in the triangulation.
The final volume mesh ready geometry was produced automatically using the surface wrapper technology. Surfaces were then automatically split, through topology detection, into three distinct domains; the primary and internal gas paths and the blade solid. The detection process identifies interfaces (surfaces common to one or more domains) and prepares them in anticipation of continuous multi-domain volume meshing further downstream in the process.

With a high quality surface mesh in place and the three distinct regions, two fluid and one solid, a polyhedral volume mesh was generated. The automation is faithful to the requirement for contiguous meshing between fluid and solid domains, and also able to perform a finite volume stress analysis to sufficient accuracy. The Finite Volume Method for Stress Analysis The finite volume methodology is more commonly used for the solution of the Navier-Stokes equations of fluid flow and heat transfer, and is common in many commercial codes. Commercial Finite Volume (FV) solver techniques for steady-state flows are frequently solved fully implicitly and therefore have no stability restriction on the time step size, unlike explicit finite element analysis (FEA). The finite volume approach presented here employs a segregated iterative solver and has been validated for wellknown test scenarios and also evaluated for consistency between different mesh types, including polyhedra. The FV method holds some significant advantages over its FEA counterpart as it has far more in common with CFD solvers than structural solvers and therefore requires similarly low memory usage for a comparable number of cells. It benefits from full parallel scalability. One seeming disadvantage is that FV stress solution requires more cells than FEA higher-order elements to properly describe bending stresses in thin members, though if compared in terms of FEA nodes, the parity is retained.

The Finite Volume Method for Stress Analysis
The finite volume methodology is more commonly used for the solution of the Navier-Stokes equations of fluid flow and heat transfer, and is common in many commercial codes. Commercial Finite Volume (FV) solver techniques for steady-state flows are frequently solved fully implicitly and therefore have no stability restriction on the time step size, unlike explicit finite element analysis (FEA). The finite volume approach presented here employs a segregated iterative solver and has been validated for wellknown test scenarios and also evaluated for consistency between different mesh types, including polyhedra.
The FV method holds some significant advantages over its FEA counterpart as it has far more in common with CFD solvers than structural solvers and therefore requires similarly low memory usage for a comparable number of cells. It benefits from full parallel scalability. One seeming disadvantage is that FV stress solution requires more cells than FEA higher-order elements to properly describe bending stresses in thin members, though if compared in terms of FEA nodes, the parity is retained.

Flow and Thermal Fields
Slices of temperature throughout blade and fluid are shown although small some cool-down is achieved by the cooling passages (see Fig:06) Indeed, if the cooling passage flow is plotted in isolation, it is seen that the fluid increases to full temperature within a very short distance of entering at the bottom of the blade.

Stress Analysis
The current turbine, overhauled by Wood Group LIT, underwent thorough examination and testing to identify long term areas of weakness of the structure. Such weaknesses can lead through minor changes in shape to inefficiencies or worse, to catastrophic failures in such turbines. It can be seen in Figure 8 that damage has been caused to shroud region, specifically resulting in a deformation affecting the external primary gas path. Figure 9 illustrates that this matches with a region of high temperature (circled). This gives a first, though not conclusive, indication of locations of potential damage. Other indicators are of the transient heating or cooling rate change in engine load. Though performed here, transient combined conjugate analysis is now made possible through this process. The effective stress, Figure 10a resulting mainly from rotational and thermal loading, supports the inference of high loading and consequent deformation at the same corner of the end-plate. Figure 10b illustrates the radial displacement resulting from the thermal field.

visit our website www.cd-adapco.com for more information about our products and services.