SOLIDWORKS Flow Simulation is a CAD-embedded computational fluid dynamics (CFD) software that enables designers and engineers to perform thermal and fluid flow analysis at design-time.

Once a Flow Simulation project is set up, the analysis is associated with the SOLIDWORKS part or assembly file, so managing design changes is simple – either changing a model dimension and re-running the analysis, or “cloning” the Flow Simulation project to various SOLIDWORKS configurations to preserve different variations of geometry.

Figure 1. Cloning an existing SOLIDWORKS Flow Simulation project.

This workflow is useful for small numbers of iterations. In other cases, where a designer wants to iterate over a wide range of operating conditions and generate some kind of performance curve, such as a pump curve for impellers, lift/drag curves for aircraft or thermal resistance curve for a heatsink, a parametric study can be used.

Creating a parametric study from an existing study allows access to three modes: “what if analysis,” “goal optimization” and “design of experiments.” Each of these has its own advantages which will be explored over the course of this article.

Figure 2. Creating a parametric study.

Case 1: Lift/Drag Ratio Curve Using “What If” Analysis

Lets discuss the case of lift and drag calculations for an airfoil. The baseline study is presented below at zero degrees angle of attack. Solution-adaptive mesh was enabled to dynamically place additional mesh refinements where needed.

Figure 3. Baseline analysis of airfoil with solution-adaptive mesh.

Once the initial project is set up, a parametric study is created using the “what if analysis” mode. Parametric studies allow varying simulation parameters (such as airspeed around the airfoil) or model geometry parameters such as SOLIDWORKS dimensions and mates.

In this case, the airfoil is placed in a SOLIDWORKS assembly with a mating scheme that allows control of pitch via an angle mate. The pitch angle is specified as a “dimension parameter” in the parametric study to allow varying the angle of attack iteratively. A range and step size is specified and a number of resulting scenarios are automatically created.

Figure 4. Angle mate input as a dimension parameter.

The “what if analysis” mode is great for “blind” analyses like this where you simply want to solve across a range of conditions, regardless of the outcome. Key results such as goals and cut plots are referenced as outputs and then the batch of scenarios can be solved.

Once the scenarios are solved, results become available and can be quickly compared across iterations, as visible in the image below:

Figure 5. Parametric study cut plots.

Numerical values can be automatically plotted against the scenarios using the built-in graphing functions. These curves can also be exported to Excel.

Figure 6. Parametric study goal charts.

To take a look at any particular scenario or “design point” in more detail, a right click allows creating a standalone Flow Simulation project from that iteration of the parametric study and opens up the capability to use the full set of results post-processing available.

Figure 7. Creating a project from a particular design point.

Note that many input variables can be specified for the “what if” analysis, but the number of scenarios will increase very rapidly and can easily get out of hand. This is something the “design of experiments” modes can help with, which we’ll look at later.

Case 2: Goal Optimization for CPU Water Cooling Block

The second mode parametric studies provide is “goal optimization.” This allows input of a specific target such as a target temperature, pressure drop and so on, and will automatically iterate the variable specified to try and achieve the target.

These studies are straightforward to setup and the results are easy to interpret. The biggest limitation is that only a single input variable can be included as part of the optimization.

Consider the case of the CPU water-cooling block depicted below. In the baseline study with 30 liters/hour of coolant flow, the predicted temperature of the chip mating surface is about 12°C above the coolant temperature.

Figure 8. Baseline study of CPU cooler.

Suppose we’d like to calculate the minimum coolant flow rate required to achieve a temperature rise of only 8°C? In this case, we’ll vary a simulation parameter (the coolant flow rate) between the range of 30 and 600 liters/hour and set a target of our maximum solid temperature to 8°C above our coolant input temperature.

Figure 9. Goal optimization input range.

Figure 10. Goal target criteria.

The goal optimization iterates until either the target tolerance or iteration limit is reached. In this case, after 10 iterations the solver indicated that a flow rate of about 75 liters/hour should produce a desired temperature rise of just under 8 degrees (visible as Design Point 10 in the figure below).

Figure 11. Tabular results for goal optimization.

Aside from the single-variable limitation of the goal optimization, there’s no way to predict in advance how many iterations it will require to converge on its target—or if it will at all—and it can be difficult to glean design trends when compared to a “what if” or “design of experiments” study.

Case 3: Design of Experiments for a CPU Water Cooling Block

While the “goal optimization” is only capable of varying a single parameter, the “design of experiments” study type comes in with multi-variable analysis. In this case, both the channel width and number of channels cut into the cooling block are varied across a pre-determined range, as depicted below.

Figure 12. Range of geometry variations for DoE study.

Whenever varying geometry (regardless of the type of parametric study) care must be taken to ensure the model geometry rebuilds properly across the range of variables specified. It’s also important that the topology of the model remains relatively consistent — especially on any faces, bodies or edges where boundary conditions or other setup conditions are specified.

It’s recommended to test the model by manually adjusting it to the “min” and “max” conditions before running the parametric study to verify that it’s able to rebuild correctly so you don’t come back to a slew of failed analyses. Such testing was performed to determine the ranges and the extremes of the two ranges are depicted below.

Figure 13. Extremes of input variable range used in DoE study.

Unlike the other study types, scenarios for a design of experiments study are not automatically created. The user specifies how many experiments to create, and the input variables are varied across the ranges. Then, after each scenario is solved, the optimum design point is estimated after-the-fact by defining an objective function.

Figure 14. DoE scenario setup.

An additional output provided by the design of experiments study is a response surface viewer, to view trends between multiple variables and their outputs in 3D. The response surface depicted below indicates that changing either variable has a nonlinear effect on pressure drop, but a mostly linear one on the predicted maximum temperature for the given flow rate.

Figure 15. Response surface viewer.

An example of extracting an optimum design point is presented below, where the temperature and pressure are set to minimize with an additional constraint on the pressure to be below a certain value. 

Figure 16. Extraction of optimal design point.

One of the great benefits of this approach is that if objectives shift, it’s not necessary to rerun the entire DoE study to achieve some new target, simply create a new optimum design point with its own objective function.

It’s worth noting that the “optimum” design points are only estimates based on the trends detected, and should be run as their own analyses to confirm that the performance matches up to their estimated performance.

Conclusion

The nature of CAD-embedded analysis tools lends itself well to fast iteration and exploration of the design space available to engineers. While it’s possible to accomplish this by manually copying existing projects, at a certain point the capabilities of a table-based iterative tool such as design studies for SOLIDWORKS Simulation or the parametric studies covered here for SOLIDWORKS Flow Simulation really start to show their benefits—whether for varying the model geometry or simulation parameters.