Improving 1D Data with 3D Computational Fluid Dynamics
Improving 1D thermo-fluid system automotive engine data with 3D CFD
By Joe Proulx, Mentor Graphics
When working with larger cooling systems that incorporate several components such as heat exchangers, thermostats, coolant pumps and different cooling cycles like oil and coolant, a 1D CFD simulation tool is the common choice for thermal analysis. Whereas when considering an automotive engine cooling system, which is rather more complex when taking into account their transient behaviors with corresponding drive cycles and system reactions with all components, becomes even more complex on any changes in flow rates or temperatures.
However, such systems can only be as accurate as the data that is supplied to them. One way to improve accuracy would be to get component data and characteristics from measurements. If however components are still in the design stage, building a prototype and measuring them can be extremely expensive. Mentor offers a direct interface by coupling CAD embedded CFD software FloEFD™ with system simulation software Flowmaster®, resulting in a CFD characterized model as a component in the Flowmaster system.
A recent paper for the SAE World Congress, (Quantifying the Accuracies between a 1D and 3D Model for Heat Transfer Calculations with an Engine) demonstrates the use of coupled 1D-3D simulation.
Description of the problem
It is well known that bringing the engine up to normal operating temperature quickly after starting is the best way to improve vehicle efficiency. The engine and complimentary components, along with the coolant and oil, all start out cold. By remanding the heat to the engine during the warm up period, efficiencies during the startup cycle of a vehicle can be improved.
A combination of 1D and 3D CFD can be utilized to determine the optimal design. Each type of simulation has its virtues, whilst 1D simulation can run long transient simulations quickly it can lack some detail. 3D simulations on the other hand can accurately simulate details of solid components but tends to be slower in regard to transients.
An engine system has many components and as a consequence, as the design develops not all the pieces are ready at once. As well as this, obtaining vital data can take time, so an efficient design process incorporating engine cooling data obtained through testing can be supplied to the 1D simulation tool for analysis. Should the component require modifications through the design cycle, testing will need to be repeated. Conversely, if testing has not been performed and all that is available is empirical data to quantify the engine, then accuracy could come into question. These types of approaches can be time consuming or inaccurate. In this instance the engine block and head were available as CAD models and the majority of the details of the 1D model were already in place making a 1D-3D approach ideal.
How CFD was used
A CAD model of the engine block and head was use for the 3D simulation. Figure 2. shows the coolant and oil flow paths. When setting up the power boundary conditions in the model, the portion of heat that is dissipated from the combustion into the cylinder walls was applied as a constant. The value of heat used for this dissipation was developed from earlier experimental data and the imposed drive cycle. Goals were set in the model to automatically capture flow versus heat transfer coefficients for each fluid. Air around the engine and head also contributed to the cooling through natural convection.
Once the base case was setup, an array of 9 models was run using a parametric study. The study was created to vary coolant and oil flow. Heat dissipations in the motor were also varied. Once the array of 9 CFD models was run, the data was then compared to empirical data according to Dittus-Boelter.
The solution and results
The 3D simulation was setup in a short time and the results from the parametric study were imported as a new component into Flowmaster for the overall system simulation of the transient drive cycle.
The graph in figure 3 below shows a significant difference in the methods used to quantify the heat transfer from the engine to the coolant. Heat transfer coefficients from the engine to the coolant had differences as high as 20% between hand calculation and 3D CFD simulations.
This shows that the overall accuracy when using 3D simulation data is far more accurate than hand calculations. The overall process of characterizing a component for a range of working parameters as done in the paper enables to system designer to evaluate any changes in the system with the same component over and over much faster than a direct 1D-3D coupling where a 3D transient simulation is the major bottleneck in the calculation time.
As shown here the detailed simulation versus hand calculation can be significantly different. Use of a validated CFD tool such as FloEFD, as opposed to a single empirical formula can greatly improve the accuracy of the data used in a 1D simulation like with Flowmaster. A single empirical formula fails to capture all of the details and differences within a detailed design. As can be seen in figure 4, there are many details in the geometry and in the results that a single formula cannot capture.
The perceived advantage of the empirical formula is that it takes less than a half hour to develop. However, in the amount of time it takes to look up the formulas, a model could be setup and a run initiated in FloEFD. From there it is only a matter of 2 days of computing time to create the data from the 3D simulation. A trade off that is well worth it for the vast improvement in accuracy.
Since FloEFD is embedded within the CAD tool, any design changes to the engine and block make it easy to capture the performance changes. If there are any changes to the engine or block, the 3D simulation could be re-run with no additional setup, thereby keeping the data current. This allows the engine and block to be developed concurrently with the 1D simulation.
As we follow along this analysis method we will see difference in the 1D modeling. A follow on paper is expected to be delivered that highlights the differences in the 1D analysis.