This project was conducted for the purpose of evaluating the appropriate applications of commercial computational fluid dynamics as an engine design tool. This was accomplished by simulating design tests normally performed by experiment. Evaluation of the prediction of flow behavior in CFD was evaluated by sequentially modeling, first, a steady state flow bench test and, second, a more complex transient motored engine. The simulations were created with the most accurate geometry representations available. Validation of the computational tool was done through comparison with experimental results.

The flow bench simulation was used to establish the validity of CFD in steady state predictions in a number of ways. First, the simulation was compared with experimental results in bulk flow measurements at two different test conditions. Second, various modeling parameters were compared with each other, such as turbulence and wall treatment, in bulk flow results and valve curtain velocity vectors. Third, several CFD software packages were compared in results of bulk flow and velocity vectors.

The second step in the evaluation sequence was the transient simulation of a motored engine. In-cylinder flow behavior was varied by the application of three different intake port designs; helical, directed and production. The simulation included full geometry representation including valve and piston motion. Flow velocity vectors average velocity and flow energy were compared with results from particle image velocimetry measurements in three planes of the cylinder normal to piston motion for each of the ports.

Results from the flow bench comparison show typical agreement within 5 percent of experiment in bulk volume flow. At the most high-flow condition however, the discrepancy between the simulation and experiment reaches a value of 13.5 percent. The reason for the difference is speculated to be a deficiency in turbulence modeling abilities. Comparisons in modeling within the same CFD code reveals a maximum difference of 3.6 percent between codes in volume flow. Of all the modeling changes, only changes in discretization order make a discernable difference in velocity profiles in the valve curtain. The comparison of different CFD codes yield a maximum difference of nine percent from the CFD average. Valve curtain velocities are shown to differ in the different CFD codes.

The second, transient simulation compared in-plane data both qualitatively and quantitatively. Velocity vector field comparisons are made using simulation results, ensemble averaged results, and individual experiment results. Comparison between simulation and experiment in velocity vector fields showed agreement in the helical port in flow behavior. Vortex location agrees in the directed port between simulation and individual experiment data. A large degree of flow disorganization causes less agreement in velocity vector properties in the production port. Quantitatively, comparisons in in-plane energy and average velocity agree with experiment to within 15 percent in six of the nine planes investigated. Discrepancies of 61 percent are found in average in-plane velocity for the directed port in the plane nearest to the head surface. Other velocity field planes are shown for informational purposes.

This project does not seek an absolute answer to the applicability of CFD to perform these design tests. The results of comparisons between CFD and experimentally obtained data are shown, the user must decide if these results justify the application of CFD to these specific tests. The determination of whether or not CFD is a useful and accurate design tool in these applications must be made on the individual's opinion of acceptable agreement with experimental results.