AOE3054 - Appendix 3 - Further Information into Experiment 7

• You rarely have the same geometry in flight that you tested. It was 15 years before Grumman had a model of the F-14 that actually corresponded to the flight article. For the F-14, as with most configurations, the configuration was "tweaked" one last time after the last wind tunnel test. The only reason Grumman eventually did have an accurate F-14 was that the AMRAAM program needed one and they were willing to pay for a new model. Other typical situations include vortex generators added during flight test programs but not addressed in original drag estimates/tests, and the gap/step correspondence between wind tunnel models and flight hardware.
• There always seems to be a disagreement over exact performance of the engine. This makes accurate flight test drag values difficult to obtain. Many, many corrections are made to the engine data. Usually there's a lot of money at stake. Both the airframe and engine must meet performance guarantees, with large penalties if they aren't met.

Another complication is the reliability of initial flight test results. The issue of flight test data accuracy further complicates the comparison between wind tunnel and flight test data. When the flight test data first arrives and the inevitable panic begins, experience shows that the flight test data system itself may be suspect, although the flight test group usually refuses to acknowledge the possibility. Two particular cases that I had some involvement with were the discrepancy between tunnel based estimates and flight test results for the loads on the vertical stabilizer of the EF-111 (upper transonic, M = 1.1, windup turns), and cruise drag of the Grumman-American Gulfstream III executive jet.

In the first case the initial flight data was over predicting the discrepancy by about 100%, and about 50% was real. A redesign of the vertical tail structure of the EF-111 was required (recall that the EF-111 differs from the F-111 by the addition of a large electronic jamming antenna placed on the vertical tail-the EA-6B has a similar pod but never flew in this speed regime). During the ensuing "fire drill" the Air Force program office was demanding daily viewgraph presentations of the results of the Grumman investigation!

In the second case the discrepancies were apparently traced to the use of a reflection plane model and two separate problems. Reflection plane models are notorious for poor drag results. The G-II reflection plane model tunnel to flight increments (which were found to agree nicely with flight data) were applied to the G-III reflection plane results. The first problem was that the G-III wing was added to the basic G-II reflection plane model, resulting in a wingspan that was too large for the tunnel. The second problem was that tunnel to flight increments for "conventional" (actually the British developed peaky) airfoils and the supercritical type airfoil used, at least in part, on the G-III scale differently between tunnel and flight (this involves the proper way to fix transition).

Cases similar to the ones described above usually aren't documented in the literature; they're not stories the participants are particularly proud of. In the end, when working with the data you begin to get a feel that you understand the problem and can account for everything. i.e. that you have "the story" (but you can't prove it scientifically).

Another famous (possibly the most famous) case involved the tunnel to flight pressure distribution on the C-141 wing. The shock in flight was much stronger and further aft than had been found in the wind tunnel. This led to trim changes and additional drag. At the time the difference was attributed to Reynolds number effects, and was used to convince congress that the a new wind tunnel was needed which could simulate flight Reynolds number using wind tunnel models (this requires that testing be carried out at very low temperatures using near liquid state nitrogen). This wind tunnel has now been built (after having been designed in large part by a VPI grad, Blair Gloss), is located at NASA Langley, and is known as the National Transonic Facility (NTF). Subsequently we tried to use this case to demonstrate the capability of computational methods to simulate Reynolds number effects. The results were disappointing, with only about half the shock movement predicted by our computational methods. When discussing our results with engineers familiar with the C-141 data, we were told that the Reynolds number was only part of the story, the other being wall interference effects. To achieve flight Reynolds number for at least marginally affordable costs the NTF is basically an 8 foot tunnel. Our biggest transonic wind tunnels are 16ft square. Thus the wall interference issue still exists in the NTF.

To conclude: Using "raw" wind tunnel data and making simple skin friction adjustments to estimate flight data may easily result in a 10% difference between tunnel and flight. The critical areas in the flight regime are almost always more complicated than first appearances. However, once you work at understanding the full story, you generally feel that you can account for the differences within a few counts. Original drag estimates made at the time that a production decision was made may not take into account a lot of the effects that occurred after the estimates were made, or did not turn out in hindsight to be wise interpretations of data. Engineering judgment is still required, and experienced engineers will be required to make tunnel to flight correlations and analysis for the foreseeable future.

The format established for the course should be used. The following comments are specific to aerodynamic testing, and may not be generally applicable.

Text (two aspects of the test report are critical)

1) In the documentation of the test enough detail must be included from the test to settle any question that comes up after the test is over. This includes the test run schedule, configuration nomenclature description, sign convention for deflecting surfaces, and the so-called "tab data" from the test. The test engineer must keep a detailed notebook during the test, and include extreme detail in the test report. Many well annotated photos should be included. Many times questions arise (sometimes years later) where the documentation is insufficient to determine with certainty exactly what happened. Few of us can remember particular test details even after a few months, and particularly when being grilled because something doesn't "look right" (this is the situation as soon as the flight test data arrives, as described above). Two typical personal examples include exact details of transition fixing, and sign convention for deflection of surfaces (at high angle-of-attack it may not be at all obvious what effect a "plus" or "minus" deflection would produce on the aerodynamic results). Since a typical test might cost several million dollars, and the data might be examined for effects that weren't of specific interest during the test, good documentation is crucial. An unfortunate, but frequent, occurrence in practice is that the time and budget expire before the test report is completed. Since it's done last, budget overruns frequently result in poor final documentation. It's best if the report can be put together while the test is being conducted. This approach can minimize the problem.

2) In the data analysis portion of the report: When writing the report provide specifics, not generalities, i.e., rather than "greater than," say "12% greater than." What do the results mean? In large organizations the test engineer will write a test report precisely documenting the test, while the project aerodynamicist will write the report analyzing the data. When writing the analysis, do not simply provide tables of numbers and demand that the reader do the interpretation. The conclusion to be drawn from the each figure must be precisely stated.

Plots and Graphs

Use real graph paper. For A size plots this means K&E² Cat. No. 46 1327 for 10x10 to the half inch, and an equivalent type for 10x10 to the centimeter. There is an equivalent catalog number for B size paper. Wind tunnel data (especially drag polars) are often plotted on B size paper. This is Albanene tracing paper. It's what's actually used in engineering work, and it's expensive. The University Bookstore stocks this graph paper. You should use it carefully, and not waste it. With high quality tracing paper, where the grid is readily visible on the back side, you plot on the back. This allows you to make erasures and also produces a better looking plot. Orange graph paper is standard, and generally works better with copy machines. The tracing paper also allows you to keep reference data on a set of plots and easily overlay other results for comparison. Remember to allow for overlay comparisons by using the same scale for your graphs.

Always draw the axis well inside the border, leaving room for labels inside the border of the paper. Labels should be well inside the page margins. Data plots should contain at least:

• reference area, reference chord and span as appropriate (include units)
• moment reference center location
• Reynolds number, Mach number, and transition information
• configuration identification

Use proper scales. Use of "Bastard Scales" is grounds for bad grades in class and much, much worse on the job. This means using the "1,2, or 5 rule". It simply says that the smallest division on the axis of the plot must be easily read. Major ticks should be separated by an increment that is an even multiple of 1, 2 or 5. For example, 10, 0.2, 50. and 0.001 are all good increments between major ticks because it makes interpolation between ticks easy. Increments of 40, 25, 0.125 and 60 are poor choices of increments, and don't obey the 1,2, or 5 rule. The Boeing Scale Selection Rules chart in included as figure 9. Label plots neatly and fully. Use good line work. In putting lines on the page, use straight edges and ship's curves to connect points, no freehand lines. Ship's curves and not French curves are used by aeronautical engineers, and some catalogs call them aeronautical engineering curves. The University Bookstore stocks at least the most common ship's curve size. As a young engineer, I was told that if the wind tunnel data didn't fit the ship's curve, the data was wrong. More often than not this has indeed turned out to be the case.

Drag polars are traditionally plotted with C_D on the abscissa or X-axis, and C_L on the ordinate or Y-axis. Moment curves are frequently included with the C_L -  curve as shown in Figure 10. The moment axis is plotted from positive to negative, also shown in the figure. This allows the engineer to rotate the graph and examine C_m-C_l in a "normal" way to see the slope.

More comments on proper plots and graphs are contained in your engineering graphics text, by Giesecke, et al. (Ref. 12). The engineer traditionally puts his initials and date in the lower right hand corner of the plot. An example of an acceptable plot is included as figure 10.

Can you use your computer to make plots and graphs? Of course. But they must be of engineering quality. To achieve this you certainly have to understand the requirements given above for hand plots, and have made enough graphs by hand to be able to identify problems in the computer generated graphs. Often it's easier to make plots by hand than to figure out how to get your plotting package to do an adequate job. Typical problems include poor scale selection, poor quality printout, and inability to print the experimental data as symbols and the theory as lines. Another problem that arises is the use of color. While color is important, it presents a major problem if the report is going to be copied for distribution. Most engineering reports don't make routine use of color - yet. A final problem frequently arises with the labeling. In reports, the figure titles go on the bottom. On view graphs and slides the figure titles go on the top. Many graphics packages are oriented toward placing the figures on the top. This is unacceptable in engineering reports.

Aerodynamic forces must be accurately measured. Initial aerodynamic testing used simple scales. Today most force measurements are made using strain gage balances. At large facilities the strain gage balances are designed, built and maintained by an instrumentation group that usually consists of mechanical engineers. However, you should understand the principles of operation. Balances must be selected for a particular test based on the physical size and loads demanded on the balance. Each balance should have a drawing indicating the size and on the same drawing you will usually find a table containing the maximum loads. Sizing and placement of the balance requires use of the pretest loads estimates made by the aerodynamics group. Typically the strain gage balance is located inside the model at a position near the aerodynamic center to minimize the possibility that the pitching moment limit will be exceeded. This can be a little challenging if the same model is to be used for both subsonic and supersonic testing, and in cases where component build-up testing may lead to situations where the pitching moment becomes large (also variable sweep wings). In some cases provision for two balances might have to be made.

Frequently aerospace companies and government agencies borrow balances to obtain the correct size. Balances are delicate and can be easily damaged if handled improperly. Their installation in models requires precision machining, and therefore, the balance is pinned in place. The portion of the model that produces loads measured by the balance is known as the metric portion of the model. Accounting for the metric and non-metric parts of the model are sometimes tricky, especially in powered model testing, where the forces over portions of the model are desired. Powered model testing is one of the most challenging areas of aerodynamic testing. Frequently the propulsion group needs more wind tunnel test time than the aerodynamics group.

Because the balance must reflect the loads accurately, care must be taken when "bridging" the balance. This means that a minimum of cables and pressure tubes should cross from the metric to the non-metric part of the installation. In some cases this means that a scanivalve or its equivalent should be placed inside the metric model to allow surface pressure and force and moment results to be acquired simultaneously. The scanivalve measures pressures from a number of orifices and allows many pressures to be transmitted to a single pressure transducer. Thus, rather than having 48 pressure tubes bridging the balance, the only lines that crosses the balance are the scanivalve, control cable, and possibly a reference pressure tube. Frequently this results in a severe temperature and vibration environment for this instrumentation, and possible problems should be considered during pretest planning.

A strain-gage balance measures forces and moments by sensing deformation of a beam-like element that has a strain gage attached to it. The balance is based on the principle that the electrical resistance of a conductor changes when subjected to mechanical deformation. A strain gage is a small, thin printed circuit type electrical resistance element which changes its resistance when elongated or compressed (Fig. 11a). This gage, which may be on the order of 1/4 inch square or smaller, which is glued onto a structural element which will be affected by the force to be measured.

A simple strain-gage balance to measure drag can be made using two strain gages on a single beam as shown in Fig. 11b. As a drag force is applied, the upstream gage element feels tension and the downstream element compression due to the bending in the beam. Since the signs of the resistance changes on the two elements are opposite, taking their difference by placing them in a circuit such that they subtract will result in a doubly strong resistance change. Any change due to a pure "upward" or "downward" force through the other beam will thus cancel out. This is important because temperature induced effects arising from thermal expansion of the material would result in misleading indications of an externally applied force.

The change in resistance can be detected by either using a constant voltage supply and measuring a current change or by using a constant current and measuring the voltage change. The resulting reading can then be calibrated by comparing the output signal under a range of test loads. Unless the gage or beam is over-stressed, the readout will be a linear function of the force applied.

Only one problem exists with the balance design in Fig. 11b, the gage output is really determined by the bending moment, which is a function of the distance between the application point of the force and the gage element location, as well as the force magnitude, instead of being a function of the force alone. A remedy to this is shown in Fig. 11c where 4 gage elements are used. Now, the circuitry can be set up to read two individual moments M1 and M2 due to the applied force and subtract them. The difference between these two moments is a function only of the force and the fixed distance between the two sets of gage elements. Since this distance is constant, the drag can be directly determined by a single circuit regardless of the location of the applied force on the beam. The circuitry for such an arrangement is shown in Fig. 11d where gage elements 1 and 2 are in tension and 3 and 4 are in compression.

A wind tunnel strain-gage balance has at least one such "bridge" to measure each force and moment. For more information on strain gage balance design see Rae and Pope (Ref. 1).

The full description of the Stability Wind Tunnel is located on the World Wide Web at the following URL https://www.aoe.vt.edu/research/facilities/stabilitytunnel.html.

1. Rae, William H., Jr., and Pope, Alan, Low-Speed Wind Tunnel Testing, Wiley & Sons, New York, 1984.
2. McLaughlin, C.B., "Experimental Aerodynamics and Configuration Design," Grumman Aerodynamics Lecture Series, May, 1985.
3. Toscano, E.J. "Aero Test Notes - Basics of What a Wind Tunnel Test Engineer Should Know About Internal Strain Gage Balances, or What a Model Designer Must Know About Internal Strain Balances and Other Things," Grumman internal notes.
4. Nark, T., "Powered Lift Testing," First Atlantic Aeronautical Conference, Williamsburg, Virginia, March 26-28, 1979. (Boeing Company Viewpoint)
5. Hedrick, I.G., "Using Wind Tunnel Test Results," Lecture XII, Introduction to Aerospace Engineering, MIT, May 3, 1982.
6. Meyer, R., "General Information on Wind Tunnel to Flight Correlation," Grumman Memo EG-ARDYN-80-150, December 15, 1980.
7. Hancock, G.J., "Aerodynamics - the role of the computer," Aeronautical Journal, August/September, 1985, pp. 269-279.
8. Braslow, A.L., Hicks, R.M., and Harris, R.V., Jr., "Use of Grit-Type Boundary-Layer-Transition Trips on Wind-Tunnel Models," NASA TN D-3579, Sept. 1966.
9. Blackwell, J. A., Jr., "Experimental Testing at Transonic Speeds," in Transonic Aerodynamics, ed. by Nixon, D., AIAA Progress in Astronautics and Aeronautics, Vol. 81, AIAA, New York, 1982, pp. 189-238.
10. E.J. Saltzman and T.G. Ayers, "Review of Flight to Wind Tunnel Drag Correlation," Journal of Aircraft, Vol. 19, No. 10, October 1982, pp 801-811.
11. "Wind-Tunnel/Flight Correlation Study of a Large Flexible Supersonic Cruise Aircraft (XB-70-1)"
• Part I, NASA TP 1514, James C. Daugherty, Nov. 1979.
• Part II, NASA TP 1515, John B. Peterson, Jr., Mike Mann, Russell Sorrells, Wally Sawyer and Dennis Fuller, Feb. 1980
• Part III, NASA TP 1516, Henry Arnaiz, John B. Peterson, Jr. and James C. Daugherty, Mar. 1980
12. Giesecke, F.E., Mitchell, A., Spencer, H.C., Hill, I.L., Loving, R.O., and Dygdon, J.T., Principles of Engineering Graphics, Macmillan Publishing Cop., 1990, pp. 591-613.
13. Ahn, Seungki, Choi, Kwang-Yoon, and Simpson, Roger L., "The Design and Development of a Dynamic Plunge-Pitch-Roll Model Mount," AIAA Paper 89-0048, January, 1989.

¹ This appendix was reviewed at Grumman under the direction of Casper Catalanotto, Group Lab Manager, Aerodynamics Test, Ground Testing Technology, Grumman Aircraft. John McAfee and Howard Jarvis contributed. Revisions were made to incorporate their comments.

² This paper is very high quality paper. With computers replace hand plotting, this paper is being discontinued by K&E. Most art supplies stores (sometimes erroneously claiming to be engineering supply stores also) don't stock good graph paper. Cheap paper will not be transparent, allowing easing tracing from one plot to another.

Figure 1 Example of the testing required to develop a modern fighter aircraft (Ref. 7).

b) Typical tuft layout on the upper surface of a wing (Ref. 1)

Figure 5. Illustration of the use of tufts for flow visualization. from Grumman Aero Report 393-81-1, October, 1981, "Experimental Pressure Distributions and Aerodynamic Characteristics of Flat and Cambered Conceptual Wing-Body and Wing-Body-Canard Models at M=1.62," by W.H. Mason.

b) drag polar

Figure 10. Examples of wind tunnel data plots. from Grumman Aero Report No. 393-82-02, April, 1982, "Experimental Pressure Distributions and Aerodynamic Characteristics of a Demonstration Wing for a Wing Concept for Supersonic Maneuvering," by W.H. Mason