Scribd
Upload
5 views1 page

Oberkampf2002 43

The document discusses the importance of a hierarchical approach to experimental measurements in computational fluid dynamics (CFD) validation, emphasizing the need for collaboration between model builders and experimentalists. It outlines a suggested hierarchy of measurements, from global forces to detailed flow-field distributions, and highlights the increasing complexity and uncertainty associated with each level. Additionally, it stresses the necessity of estimating both random and bias experimental errors to ensure accurate validation of CFD models.

Uploaded by

Abit Bronja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views1 page

Oberkampf2002 43

The document discusses the importance of a hierarchical approach to experimental measurements in computational fluid dynamics (CFD) validation, emphasizing the need for collaboration between model builders and experimentalists. It outlines a suggested hierarchy of measurements, from global forces to detailed flow-field distributions, and highlights the increasing complexity and uncertainty associated with each level. Additionally, it stresses the necessity of estimating both random and bias experimental errors to ensure accurate validation of CFD models.

Uploaded by

Abit Bronja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

W.L. Oberkampf, T.G.

Trucano / Progress in Aerospace Sciences 38 (2002) 209–272 251

close cooperation and communication between the quantities. The experimental uncertainty increases as
model builder and the experimentalist; sometimes they one proceeds down this list, probably at a factor of two
are the same person. for each level of the hierarchy. With the recent
development of experimental techniques such as particle
Guideline 5. A hierarchy of experimental measurements imaging velocimetry (PIV) and planar-laser-induced
of increasing computational difficulty and specificity florescence (PLIF), we believe there is a significant
should be made, for example, from globally integrated increase in the quantity of available flow-field velocity
quantities to local measurements. measurements. The quantity of data, for example, over a
large volume surrounding a body, permits much more
As one moves from global to locally measured stringent tests for CFD model validation than simply
quantities, the challenge to the computationalists and measurements at individual point locations.
the experimentalists increases significantly. In wind- The last recommendation concerning the hierarchy of
tunnel experimentation for a flight vehicle, the following measurements is that in a validation experiment one
hierarchy is suggested: should, if possible, make measurements at multiple
levels of the hierarchy. That is, do not design the
* Flight vehicle forces and moments. experiment with the philosophy that only one detailed
* Control-surface forces and moments. level of measurements will be made. The more compli-
* Surface-pressure distributions. cated the flow field or the physical processes taking place
* Surface heat flux, shear stress, or both. in the flow field, the more important this recommenda-
* Flow-field distributions of pressure, temperature, and tion becomes. For example, on a complex turbulent,
velocity components. reacting flow, do not just measure surface heat flux over
* Flow-field distributions of Reynolds stresses. a portion of the body. Also measure flow-field tempera-
tures and surface pressures over a different portion of
The ordering of difficulty in the above hierarchy the body. Flow-field visualization and surface-flow
indicates that each lower level, i.e., higher level of detail, visualization can also provide valuable additional pieces
is either the spatial or temporal derivative of the level of information. With sophisticated postprocessing cap-
above it, or it is a subset of the level above it. In other ability, the CFD solution can be used to simulate the
words, the integration or selection of a larger set is a experimental flow-field visualization and surface-flow
powerful mathematical smoothing process. Thus, there visualization. For example, the computed flow-field
is a relationship between this hierarchy and the levels of solution can be used to compute a Schlieren or
credibility in validation activities. That is, in the interferometer photograph that can then be compared
‘‘process of determining the degree to which a model is with the experimental photograph.
an accurate representation of the real world,’’ one must
specify what physical quantities, or system response Guideline 6. The experimental design should be con-
measures, have been validated. For example, validation structed to analyze and estimate the components of
of body normal force does not imply that surface heat random (precision) and bias (systematic) experimental
flux to the vehicle has been validated to the same degree errors.
of accuracy. There are two separate reasons for this
statement. First, the physical modeling fidelity that is The standard technique for estimating experimental
required to predict these two quantities is remarkably uncertainty in wind-tunnel data has been developed
different. For example, body normal force on many over the last ten years by members of the AGARD
geometries can be computed fairly accurately with Fluid Dynamics Panel [8]. The standard procedure,
inviscid flow, whereas the difficulty of predicting surface although not widely used in wind tunnel facilities, is
heat flux in a turbulent boundary layer is well known. well documented in a recent AIAA standards document
Second, the computational grid size that is required to [13] and also in the text of Coleman and Steele [75].
predict these two quantities is strikingly different. For We believe it is the minimum level of effort required
example, consider a steady, compressible, laminar, for uncertainty estimation in validation experiments.
attached flow over a delta wing at a low angle of attack. The standard technique propagates components of
To achieve the same level of computational accuracy for random and bias uncertainty through the entire
body normal force as compared to surface heat flux, one data-flow process. The technique estimates these com-
would need approximately a factor of 100–1000 times ponents and their interactions at each level of the
the number of grid points for the heat-flux computation process, from the sensor level to the experimental-result
as compared to the normal-force computation. level. As with all techniques for estimating experimental
The predictive difficulty for a code that is illustrated uncertainty, the ability to estimate random uncertainty
by the above hierarchy is also very similar to the is much better than the ability to estimate bias
difficulty in experimentally measuring each of these uncertainty.

You might also like