Evaluating Thermal Design of Construction Vehicles

April 1, 2016

Design and evaluation of construction equipment and vehicles constitute a very important but expensive and time-consuming part of the engineering process, due to a large number of prototype variants and low production volumes associated with each variant.

Engineers from Exa Corp. and Charles Machine Works collaborated to investigate an alternative approach to hardware testing. Because of the enormity and complexity of the vehicle design process for off-highway vehicles, the researchers limited the scope of their current study to thermal evaluation of vehicle components only. A CFD simulation provides the surface temperatures of the entire vehicle and the flow domain. However, the focus was on two particular areas that address critical and challenging thermal issues: electronic control units (ECUs) and hot air recirculation which occurs when the vehicle is stationary or at very low speeds and when engine cooling and airflow are provided by the fan only.

Electronic components are challenging since their maximum temperature requirement is relatively low while the thermal environment can be very hostile as a result of heat sources in close proximity. For hot air recirculation, it is possible due to a combination of vehicle geometry and seals for the hot air exiting the heat exchangers to recirculate back to the inlet of the heat exchangers, leading to potentially catastrophic temperature excursions and failure of components.

For the CFD simulations, performed on a Ditch Witch RT120 Quad vehicle, a coupling is implemented between the Flow Solver PowerFLOW version 5.0c and the Thermal Solver PowerTHERM version 11.0.4 that accounts for conduction and radiation effects.

One novel aspect of the simulation is extensive use of solid meshing for the thermal model in order to model conduction extremely accurately, which is very important for the ECUs since they are located in a low airflow environment and conduction is the most important mode of heat transfer for that location.

This operating environment is thermally challenging since the cooling is provided by the engine-driven fan alone and the vehicle is stationary. The flow is strongly dominated by forced convection near the fan and natural convection away from it.

Validation studies where PowerFLOW simulation results were compared to experiments have been documented in prior SAE technical papers. Good agreement was observed between simulations and experimental data for both forced and natural convection flows.

Simulation setup

The flow solver is a general-purpose CFD code that uses the Boltzmann Equation to solve for the velocity and pressure field. The turbulence model employed in the flow solver is conceptually similar to a large eddy approach to turbulence modeling. Recent improvements in modeling the temperature scalar equation have further enhanced the accuracy of the temperature calculations in the flow solver.

The simulation methodology involves a combination of the flow and thermal solver. The flow solver calculates the flow and temperature fields in a volumetric flow domain, and the thermal solver calculates the temperature in the solid surfaces by taking into account radiation from the surfaces, conduction within the solid surfaces, and convection at the surface.

The current study is, as far as the authors know, the first of its kind that looks into a stationary vehicle with the fan on and is able to accurately simulate the flow and temperature field as evidenced by good comparison with test data. Furthermore, the use of a large eddy-like flow solver with fully transient formulation is not something that has been attempted before for an off-highway vehicle in idle.

Since one of the objectives was to evaluate hot air recirculation of the heat exchangers, the geometry of the vehicle hardware between the radiator and the fan was modeled “as is” from CAD and no effort was made to close out small air gaps between components.

The vehicle has four heat exchangers namely the radiator, charged air cooler, and the left and right oil coolers that are each modeled by heat exchanger characterization data. The airflow through the heat exchanger cores on the air side are modeled using the porous media feature of the flow solver. The resistance coefficients for the porous media were based on measured pressure drop versus flow rate data.

The vehicle was stationary and all the cooling was provided by the rotating engine fan. The rotating fan was accurately modeled using the sliding mesh capability of the flow solver.

Thermal model

For the thermal model, accurate inputs of thermal material properties for different vehicle components are required. Material property inputs are thermal conductivity, density, specific heat, and surface emissivities.

The heat sources were modeled as prescribed temperatures that were obtained from supplier-provided test data. To accurately model conduction, 2D shell elements and 3D tetrahedral elements were used for the entire thermal solver model.

The extensive use of solid elements is a novel feature that was done to account for the conduction paths in a more complete manner for a complex 3D geometry and hence accurately model thermal conduction for the solid parts. For earlier studies using coupled thermal simulations, thermal conduction was modeled using 2D shell elements with prescribed part thicknesses. This approach worked well for thin parts such as thermal shields with relatively simple conduction paths; however, for this study, not all thermally critical parts could be approximated as thin parts and therefore 3D solid elements were used for the thermal solver.

For this study, the critical parts are the ECU units and batteries located under the step. The ECU unit and battery are shielded from the heat coming from the inboard side by seals, insulation, and frame rails that separate the hot inboard side and the ECU compartment.

The insulating layer that separates the hot inboard side from the compartment is not thin and has complicated conduction paths that make it difficult to model conduction accurately using 2D elements alone. Therefore, 3D tetrahedral elements were used to model conduction.

Simulation results and test comparisons

The vehicle is in the idle regime wherein the fan is rotating but the vehicle is stationary with no wind velocity assumed in the simulation or observed in the test environment. The experiments were performed in a test chamber at Charles Machine Works. The temperature measurements were made using thermocouples, which were accurate to within 2°F (1.1°C).

It is important to validate the results and gauge the accuracy of the simulation. The temperature distribution in a vertical plane upstream of the radiator provides a good visualization of the pre-heat that is caused by recirculation. Note that the ambient air is 104°F (40°C). All the pre-heat was found to be a result of recirculation of hot air.

Good agreement is observed between the experiments and that of the simulation. The difference in the temperatures is close to the experimental uncertainty of the measurements, which is estimated to be 5°F (2.8°C). Given other input uncertainties, the difference of less than 10°F (5.6°C) between simulation and experiments is “good.”

The difference of 4°F (2.2°C) of the coolant temperatures between the experiments and simulation are within the experimental uncertainty. Note that the fan rpm, which is assumed as constant in the simulation, fluctuates by up to 4%, which leads to a 5°F fluctuation in the coolant exit and entry temperatures as per sensitivity studies with Flow Solver.

For the temperature distribution in a horizontal plane upstream of the charge air cooler, good agreement is observed between the experiments and that of the simulation with the exception of the bottom left location where the pre-heat is under-predicted. Geometry mismatches, in particular with the seals, are a possible cause of this inconsistency. Good agreement is also observed for the charge air cooler inner air temperatures and heat rejection.

Within the ECU compartment, which is located under the step near the hydraulic tank, the velocity levels are quite low, as shown along an x-section; therefore, temperature distribution is dominated by conduction since convection levels are low on account of low velocities.

Researchers observed that the high temperatures are on the inboard side of the ECU compartment and the compartment is heated primarily from the hot air temperatures in the inboard side that is transported via solid conduction through the frame rails.

They also observed a temperature gradient in the y-direction of the compartment. The temperature drops by 12°F (6.7°C) as it moves outboard 450 mm (17.7 in) in the plus y-direction. The temperature gradient observed compares well with the test data, where a temperature gradient of 1°F (0.6°C) was observed over a distance of 40 mm (1.6 in).

For temperature distribution in a y-section, the test thermocouple measurement was 150°F (66°C) and the simulation reported a temperature of 156°F (69°C). Therefore, good agreement was observed, taking into account the uncertainty of the thermocouple location and that temperature gradients were observed in the compartment air.

The average surface temperature distribution for ECU1 (upper) at 166.9°F (74.9°C) and ECU2 (lower) at 161.7°F (72.1°C) reveal that both ECU components meet the requirements of maximum temperatures of 180°F (82°C). The test thermocouple measurement on the surface of ECU1 was 162°F (72.2°C). Therefore, good correlation is observed between test data and simulation for surface measurements.

In addition to the promising results, the total time required to set up and run the simulations is very reasonable and could reduce time and cost of the engineering design process for off-highway vehicles.

This article is based on SAE Technical Paper 2015-01-2888, doi:10.4271/2015-01-2888, by Mukutmoni, D., Donley, T., Han, J., Muthuraman, K. of Exa Corp., and Campbell, P., Mertz, T. of Charles Machine Works.

---