|INSIGHT: Mind the Gap--Diagnosing Discrepancies Between CFD Predictions and Experimental Data
How do you handle situations where your CFD predictions and experimental data are pointing you in two different design directions? In this month's issue of The Flow, we sit down with Bob Ni to discuss the causes of these discrepancies and a simple diagnostic to help you find your way. Prior to founding ADS, Bob spent nearly 30 years at Pratt & Whitney leading turbomachinery CFD in support of compressor and turbine design.
FLOW: Bob, how do you diagnose unusual discrepancies between CFD predictions and rig data?
BOB: As we all know, CFD predictions rarely, if ever, perfectly match experimental results. Some discrepancy is always to be expected since it's a modeling tool; you just want to make sure that the CFD is accurate enough to consistently discern design improvements from design mistakes.
That said, when a discrepancy arises where the CFD and rig data are pointing me in two different directions, I find it helpful to run through a mental checklist to help guide my actions:
- Do I have the right geometry?
- Have I modeled the important features?
- Are the computational boundaries properly defined?
- Have I hit a CFD limitation?
- Could the experimental data be wrong or misinterpreted?
FLOW: Let's delve into these checklist items a bit. What do you mean by "having the right geometry"?
BOB: Well, the first step is to always make sure the geometry you're analyzing is the same as what you've tested. Sounds simple but you have to be careful. One common error is to overlook the distinction between "cold" geometry data used for manufacturing and "hot" geometry used for predicting aerodynamic performance. Hot geometry assumes the turbomachine is at operating conditions and accounts for small differences in geometry due to thermal expansion. These small differences can directly impact critical design parameters such as tip clearance that can materially change the CFD predictions. So as a first step, make sure you're using the right geometry --the hot geometry--in your CFD analysis.
FLOW: As a second step, you recommend verifying that important features have been modeled. Is this an obvious statement?
BOB: Actually not--it is very easy to overlook features that seem insignificant but can contribute to material discrepancies between predicted and measured data. Two classic examples: tip clearances and endwall gaps. So as a second step, be sure that these types of features are always accounted for in your analysis.
FLOW: Computational boundaries are next on your list. What are some of the common pitfalls on this front?
BOB: First you should make sure your inlet and exit conditions are correct. I can't tell you the number of times a negative sign gets missed on rotational speed or flow angle because of directional dyslexia! Often it's also important to make sure that your upstream and downstream boundaries are set sufficiently far enough away from the leading and trailing edges to ensure clean CFD results. For example, we typically like to make sure that the computational boundaries are set at least 25% of the axial chord length away from the leading and trailing edges. So step three should be to reverify your computational boundaries.
FLOW: Step four is to consider the limitations of your CFD package. What types of "CFD limitations" are typically encountered?
BOB: It's critical to have a good feel for the inherent limitations of your CFD package. Some typical limitations relate to meshing and turbulence modeling. For example, be sure to understand how the mesh topologies or densities offered in your CFD package may be limiting your ability to reveal important flow phenomena in the interest of turnaround time, or how sensitive your solver is to mesh skew resulting from a particular design configuration. Consider also what the limitations of your CFD package are for handling different flow regimes or turbulence modeling.
FLOW: Lastly, you recommend checking the experimental data. How can experimental data be misinterpreted?
BOB: In a world where the experimental data is the "gold" standard, it's important to recognize that it too can occasionally suffer from inaccuracies. So if you're satisfied that you've eliminated all other possibilities, be sure to allow for the possibility that your rig data may be off. For example, instrument location and interference may skew gathered data, or the time scale of a blowdown facility may not be sufficient to capture steady state effects (such as heat conduction). Or the calculation or a parameter may be very different from the way your CFD package calculates it. For example, area-averaged vs. flow-averaged calculations.
FLOW: How would you summarize your approach to our readers?
BOB: When unusual discrepancies arise between the CFD predictions and experimental data, they can invariably be tied back to five root causes: the wrong geometry, missing features, incorrect boundary conditions, CFD software limitations and/or misinterpreted experimental data. By working through the five simple questions outlined in this article, you should be able to systematically "mind the gap" and minimize errant decision-making.
FLOW: Thanks, Bob.
BOB: My pleasure.
CASE STUDY: Analyzing the NASA Glenn Transonic Flutter Cascade with Code Leo and Code Wand The NASA Glenn Research Center Transonic Flutter Cascade is a linear cascade wind tunnel capable of operating at transonic Mach numbers. In this case study, Richard Elifritz and Vince Capece from the University of Kentucky conduct 2-D analysis using Codes Leo and Wand. <more>
TECHTIPS: Visualizing Results with ParaView Animations
Animations can be used to explain results that are often difficult to convey using static images. ParaView offers an easy and intuitive way to create animations from any ADS restart file. <more>
Welcome to The Flow
Welcome to The Flow, a newsletter for monthly insights on turbomachinery CFD published by AeroDynamic Solutions, Inc.
Each month we'll spotlight a topic of interest, discuss a case study and/or provide useful pointers about how to get the most out of the ADS CFD system.
You are receiving this email because you or someone else you know thought you would be interested. To unsubscribe please click here. We value your privacy.