Co-Simulation Analysis
As the latest generation engines require various heat exchangers for engine oil, transmission oil, alternator and passenger cabin heater there are usually more than one outlet from the engine-cooling jacket. BMW's engineers determined that the two flow regimes, engine and external system, are not independent. The 3-D computation must use the global flow conditions as boundary conditions for the detailed local flow calculations and the local pressure drops must be used in the global flow calculations. Therefore, an iterative process is required for a solution of the overall system. In general, a single iteration is insufficient for convergence and so the process must be repeated.
The data transfer was originally performed manually. The shortcomings of this process, slowness, cumbersome and error prone, led to the development of a programme interface, which improved the productivity but still required setting set-up for each case. The interface was built with standard FLOWMASTER2 utilities Command Line Interface and Electronic Data Interchange Format, and STAR-CD user subroutines.
BMW's engineers applied the method to the cooling system of a six-cylinder diesel engine. The cylinder head and cylinder cooling-jacket was modelled with STAR-CD and the external cooling circuit with FLOWMASTER2. There were three connections between the two codes - water pump, cylinder head and cylinder-jacket.
The systems were analysed independently to derive the boundary conditions, mass flow rates, for the co-simulation analysis. Good agreement between the boundary mass flows was achieved after three iterations. Convergence, where the boundary mass flows from both codes were within the specified tolerance, was achieved after 12 iterations.
Results
The figure below shows the percentage change in pressure boundary conditions versus iteration count during the iterative analysis. As each data transfer takes place the two computations converge. The upper line represents the cylinder head - red for STAR-CD and orange for FLOWMASTER2. It can be seen that the changes are large for the first two couplings but converge quickly. Also, the converged STAR-CD values show about 20% change from the initial estimate. The lower line representing the cylinder jacket shows a similar trend but the final change is much larger - about 40%.
Percentage change from initial pressure boundary conditions. As each data transfer takes place the two computations converge. Cylinder head: orange - FLOWMASTER2, red - STAR-CD. Cylinder-jacket: light blue - FLOWMASTER2, medium-blue - STAR-CD
The results, with coupling, were very different from the original values without coupled analyses. The mass flow at the water pump was 4% lower than assumed. However, the flow distribution between the cylinder head and jacket was very different. The split was originally predicted to be cylinder head - 83%, cylinder-jacket - 17%, but the results of the coupled solution showed it was cylinder head - 69%, cylinder-jacket - 31%. This result was unexpected and as a result the design was modified to increase the cylinder bore centres allowing increased cylinder-jacket size.
Conclusion
BMW Motoren GmbH, Steyr, has combined 3D with 1D analysis to analyse the cooling system of a six-cylinder automotive diesel engine. They found that results of the co-simulation were an improvement over the independent analyses. The results showed that the original assumptions were inaccurate and that original design had been
sub-optimal.
|
]
