Cool solutions off road for Caterpillar

Starting with a CAD model of the test rig, a trimmed cell mesh (shown in the figure) was generated using the underhood expert system module es-uhood. The ventilation airflow field in the test rig and the convective heat transfer coefficients for the solid surfaces were determined using STAR-CD. An initial parametric study on importance of the buoyancy force in the thermal-fluid calculations revealed that the effect of density variations on the overall flow and temperature fields was negligible. Thus, the ventilation airflow field was simulated as a steady incompressible flow using the high-Re k-e turbulence model with logarithmic wall functions. As the most basic two-equation model, k-e model is believed to provide a reasonable approximation of the time-averaged flow distribution over the surface of the engine and its components in the test rig. A set of transient calculations were also studied to investigate temperature fluctuations observed during the experiments and to assure that the calculated flow field is steady with no oscillations. The calculations were performed on a Linux cluster.

As the examples of results obtained with the CFD model, the ventilation airflow field and temperature distributions are shown in the figures above on a vertical plane through the enclosure front inlet. The results indicate that the most significant pressure drop takes place near the inlet and outlet restrictions. Consistent with the experimental observations, the results also reveal a well mixed flow inside the enclosure with no significant difference in component temperatures for different ventilation inlet locations.

The comparison of the experimental and CFD model predictions for pressure drop through the test rig is shown in figure 3 as a function of airflow rate. A good agreement for such “system restriction curves” is the first indication that CFD model captures the flow field accurately. For the configurations studied so far, the comparisons for ventilation air temperatures throughout the enclosure are also consistent with the experimental values when the surface temperatures are specified as the boundary conditions.

A network flow model of the engine with reduced complexity is also built using commercial software Flowmaster, and it is currently being coupled with STAR-CD model. The network flow model consists of 1-D descriptions of thermal subsystems including the coolant, oil, and air loops combined with a lumped parameter approach to characterize the thermal interactions between them through the engine structure. i.e. the heat generated during combustion is considered to be transferred to various discrete surface points on the engine using specified conduction paths. In this simulation, the radiator is simply modeled as a sink with known characteristics (constant flow rate and inlet temperature).

During the 3-D and 1-D co-simulation, the ventilation airflow field inside the engine enclosure and the rate of heat transfer between the engine and air are determined with STAR-CD. The coupled model requires flow rates and inlet temperatures as the boundary conditions in the ventilation air and coolant loops and oil pump speed in the oil loop to account for overall energy balance and predict the engine component temperatures. In the STAR-CD model, these predictions are prescribed as surface temperature boundary conditions for various engine components and enclosure walls. The results of the 3-D CFD analysis are, in return, provided back to the 1-D model to improve component temperature predictions by modifying the air flow paths and heat transfer coefficients between the engine components and ventilation air. The typical values of estimated heat transfer coefficients between the engine components and ventilation air are found to vary in the range from 10 to 50 W/m2-K.

The results to date indicate that the temperatures and distributed heat rejection rates can be estimated within reasonable accuracy when 3-D and 1-D models are used in combination.