STAR-CCM+ - An Elegant Engineering Solution to Engine Conjugate Heat Transfer

Richard Johns, Automotive Director CD-adapco

Structural Integrity of major engine components is the one of the most fundamental of design criteria that is considered by powertrain engineers. The loads on the cylinder head and block are complex and include assembly and operating loads, residual stresses and, especially towards the top of the block and the cylinder head, thermal loads. Indeed, fatigue failure resulting from thermal cycling has been the cause of many engine component failures in the past and the increase in thermal loading from the recent trends in downsizing and engines of high specific output will keep this in focus for the foreseeable future.

Since the early 1970s, Finite Element calculations of temperatures and stresses in engine structures have been undertaken by major OEMs throughout the world. Accurate calculation of the temperature field is critical to determining the stresses and this, in turn, depends upon the accuracy of a computational mesh generated to fit the geometry and the boundary conditions that are applied on both the gas and coolant surfaces. Determination of the gas-side boundary conditions is beyond the scope of this article but, suffice it to say that methods are available in STAR-CD and involve detailed calculations of the in-cylinder flow and combustion.

In the early days of engine structural temperature calculations, coolant-side boundary conditions were determined using empirical formulae based on Nusselt Number – Reynolds Number relationships and with significant experimental input under appropriate conditions to ensure their validity. In the early 1990s as CFD methodology matured and computers became faster (and cheaper!), coolant flow calculations became commonplace. Initially, meshes were generated from drawings and could take many weeks of laborious work. As good 3D CAD data became available and robust automeshing techniques were developed, this task became automated with a step change in usefulness as the analysis was able to not only keep pace but to lead the design.

Although the flow within the coolant passages is of interest in its own right, it is the structural temperatures which are of primary interest; here, analysis developments have followed a more leisurely pace. The multi-code, thermal boundary condition exchange process shown schematically in figure 1 has been used widely for many years to deliver the structural temperatures. Separate programs are used alternately for the flow and structural temperature solutions with to-and-fro mapping of heat transfer coefficients and temperatures between the programs. After a few cycles of this process, the solution usually converges and the process is terminated.

Although this is a pragmatic engineering solution which is still used widely, there are a number of potential disadvantages: firstly, it is not a simultaneous solution and as the temperatures of the solid and coolant and the flow conditions are coupled, care must be taken to ensure both local and overall convergence of the heat flux. Secondly, it is time consuming building 2 meshes for 2 different programs – often different types of meshes for finite-volume and finite-element based solvers and the multi-program solution process. Finally, and potentially of most importance, local boiling can occur, especially in regions of the cylinder head, and proper inclusion of this phenomenon requires a boundary condition that is dependent on both the local fluid conditions and the structural temperature; it also requires tracking of the vapour within the flow passages. Clearly, there is an opportunity to improve on the existing process and the remainder of this article describes the STAR-CCM+ solution that overcomes all of the above deficiencies.

STAR-CCM+ has the ability to generate meshes in multiple domains and if the domains are connected and separated by common boundaries, then the meshes at these boundaries are conformal with one-to-one connectivity. In the context of engine coolant flow and thermal analysis, different parts of the analysis model can be assigned to different domains such that the coolant flow passages, lube oil circuit and different parts of the structure (such as aluminium, cast iron and gasket materials) are delineated and can be assigned appropriate properties.  Figure 2 shows a section through an engine structure with the head, block, gasket and coolant passages clearly identified. Also shown is the use of prism layers adjacent to boundaries and introduced by the user to enhance accuracy in the application of boundary conditions and to resolve gradients close to the surface.

Conjugate heat transfer problems are addressed in STAR-CCM+ by solving the thermal energy equation in all solid and fluid domains simultaneously whist the fluid flow is only considered in those for which it is relevant. In addition to the traditional convective flow boundary conditions the heat transfer between the solid and the fluid incorporates a model for sub-cooled nucleate boiling which, in turn, depends upon the local pressure, fluid and structural temperatures and fluid properties. In addition to enhancing the rate of heat transfer under boiling conditions, vapour is also generated and a transport equation for vapour fraction is also solved.

There are a number of advantages to this approach compared to the multi-code solution methodology described earlier:

1. Faster: it is a “one pass”, single code methodology – see figure 3, in which there is no iteration between codes
   
   
2. More Accurate:  A simultaneous solution to the coupled fluid-structural flow and heat transfer is obtained. Furthermore, it is the only way in which boiling can be incorporated properly.
   
   
3. Integrates into Existing Processes: The way in which gas-side boundary conditions are applied is identical to existing methods – instead of mapping to an FE surface the data is mapped to a STAR-CCM+ mesh.

 

Validation of the model has been carried out with reference to experimental rig data and good agreement was found for the heat flux for flow, pressure and surface temperature conditions typical of those found in engines. Figures 4 and 5 show some typical results from the same multi-cylinder engine shown in figure 2. Run time for this calculation was 5 hours on 10 cpus but overall process time was significantly reduced compare to the original multi-code approach.

Many OEMs are now assessing the STAR-CCM+ solution described above and anticipate both process efficiency and accuracy improvements through adopting this technology. Clearly, this is but one application in one industry and this methodology has potential benefits across a wide range of analogous problems elsewhere.