One of the more urgent questions facing American aeronautical engineers in the 1920s was how to reduce the drag of radial engines without degrading their cooling. Soon after he end of World War I, the navy had become convinced that the air-cooled engine offered a more practical solution to its aircraft power-plant problems than did the heavier liquid-cooled engine-with its water jacket, radiator, and gallons of coolant-favored by the army. The jarring confrontations of naval aircraft with arresting gear on aircraft carriers resulted in too many cooling system maintenance problems at sea (e.g., loose joints, leaks, and cracked radiators). However, subsequent experience also made it clear to the navy's Bureau of Aeronautics that existing air-cooled designs wasted considerable power: projected into the external airstream for cooling, the finned cylinders of the radial engine caused high drag. Navy engineers attempted to reduce this drag by putting a propeller spinner (a rounded cover) over the hub and covering the crankcase and inner portions of the cylinders with a metal jacket, but this left the outer ends of the cylinders still jutting into the airstream.1
In June 1926 the chief of the Bureau of Aeronautics asked the NACA to determine how much a cowling could be extended outward over the cylinders of the radial engine in order to reduce drag without excessive interference with cooling. Less than , year later, during a technical session of an aircraft manufacturers' conference at Langley Memorial Aeronautical Laboratory, representatives from industry also asked the NACA for help in understanding the effects of cowling on the performance and cooling of radial engines.2 The NACA responded to these requests by authorizing its laboratory, first, to conduct a free-flight investigation of the effects of various  forms of cowling on the performance and engine operation of a Wright Apache (borrowed from the navy) and, second, to prepare a systematic program of cowling tests in its Propeller Research Tunnel (PRT), a brand new facility that made it possible for the first time anywhere to test full-size propellers and other aircraft components in a wind tunnel.3
The results of this research program are well known. In 1929 NACA Langley won its first Robert J. Collier Trophy, an annual award presented by the National Aeronautic Association for the year's greatest achievement in American aviation, for the design of a low-drag cowling.4 By the mid-1930s the laboratory had designed a family of streamlined cowlings that not only reduced drag dramatically but actually improved engine cooling as well, an accomplishment that confounded the previous engineering intuition that had stuck those finned cylinders directly into the airstream. What is not very well known, however, is the history of the Method of the NACA's successful cowling research and, more specifically, the fact that the engineers at Langley who used that method so well in the late 1920s and early 1930s eventually met and had to overcome what can only be described as an experimental impasse, a position from which there seemed no empirical way out. That history is the subject of this chapter.
The primary method employed by the NACA engineers in their cowling research was experimental parameter variation-"the procedure of repeatedly determining the performance of some material, process, or device while systematically varying the parameters that define the object or its conditions of operation."5 When a complex research problem needs practical solution, and hypotheses are more scattershot than pinpoint because complex understanding is still a distant goal, this technique systematizes the pragmatic researcher's only real choice for a course of action: a combination of brain work, guesswork, and trial and error. By observing the effects of slight changes made one at a time in planned, orderly sequence, he can add progressively to his knowledge about the actual performance of whatever he is investigating. Seeking effects now and saving causes for later, he uses what he does know, circumvents what he does not know, and discovers what will work.
The method is ancient. Greek military engineers varied the parameters of full-scale machines to find the most effective dimensions for their catapults.6 During the Industrial Revolution, engineers used the method to explore the performance of new construction materials and steam engines.7 The success of the first powered airplane in 1903 followed application of....
....the method's fundamentals by the Wright brothers while testing airfoils in their wind tunnel.8 The success of Langley laboratory's own airfoil research in its Variable-Density Tunnel, discussed in the two previous chapters, also hinged on parameter variation, the method has been used for so long by so many different types of engineers precisely because it permits solution of a complex problem without a complete understanding of all aspects of the problem.9
The growth of the method of Langley's cowling research from 1926 to 1936 and beyond can be divided into four stages: (1) definition of the cowling's parameters, ending in 1929 with the public announcement of a successful low-drag design; (2) 1929 to 1931, encompassing an important series of engine placement and free-flight cowling tests that resulted in a strong identification throughout the NACA with the empirical method; (3) 1931 to 1934, when the laboratory began by outlining a new three-pronged experimental attack on cowling and cooling problems, but ended in an impasse with that attack stalled; (4) 1934 to 1936 and beyond, when a more analytical approach to cowling research began to emerge out of this stalemate to answer some of the basic questions that the empirical approach of the preceding stages had left unanswered. Experimental parameter variation had led to results. Practical use had been made of observed performance effects; now it was time to search beneath them for those causes that had been postponed for later. It was time to go after that distant goal of complex understanding.
LMAL engineer Fred Weick, he former employee of the navy's Bureau of Aeronautics who had become principal designer and head of the Propeller Research Tunnel, was already very familiar with parameter variation when he used it as the basis of the first-stage attack. As a senior engineering student at the University of Illinois, Weick had based a paper on variable-pitch propellers on data from "tie first aerodynamic study under NACA auspices" to employ the method of experimental parameter variation-data reported by professors William F. Durand and Everett P. Lesley from model air propeller tests in the Stanford University wind tunnel.10 Later, one of the first things Weick had done after joining the Bureau of Aeronautics in 1924 was "to work out a simple system of blade-element analysis using only a single element ... but obtaining the airfoil lift and drag characteristics by working the analysis backward" from the Durand-Lesley propeller data. In 1926, Weick reported the details of the method in NACA Technical Notes (TNs) 235 and 236.11
Recognizing that he should extend the cowling investigation well beyond the range of immediate interest, Weick pinpointed the extremes. Obviously, one extreme was a bare engine with no cowling at all; everyone who....
....knew anything about aerodynamics assumed that it would have maximum cooling, but maximum drag as well. The value of the other extreme - enclosing the engine completely - no on had anticipated because that form seemed to exclude all possibility of air cooling. For smooth flow around the exterior of the cowl, Weick's team of engineers in the PRT modeled an engine nacelle on the best available airship form. Then the amount of cowling was systematically varied from extreme to extreme, and ten different cowlings for experimentation resu1ted.12
The test program proceeded easily enough - its goal being a cowled engine that would be cooled just as effectively as one with no cowling whatsoever. The PRT team mounted the Apache's J-5 Whirlwind engine in the tunnel, measured the cooling effectiveness of each of the ten cowlings, and investigated their effects on propulsive efficiency. Each experimental shape underwent numerous, systematically planned variations. With the help of Elliott G. Reid, the head of the Atmospheric Wind Tunnel section, who had been studying the effects of Handley-Page wing slots, the team designed a cowl that brought outside air in and around the engine via a slot....
....at the center of the nose. The potential of a complete cowl then began to look more enticing. Researchers had to modify the cooling air inlet several times, and had to install guide Vanes or baffles to control the air in its passage for a more efficient heat transfer. They also had to design an exit slot that released the air at a slightly higher velocity and lower pressure than it had entered the cowling with, but they finally obtained satisfactory cooling with a complete cowl (called "no. 10") that entirely covered the engine and used slots and baffles to direct air over the hottest portions of the cylinders and crankcase.13
To everyone's surprise, the no. 10 cowling reduced drag by a factor of almost three! The results of this first portion of cowling tests at Langley were so remarkable that the NACA made them known to industry at once. In November 1928 the Committee published Technical Note 301, "Drag and Cooling with Various Forms of Cowling for a 'Whirlwind' Engine in a Cabin Fuselage," by Fred Weick. In it, Weick argued that use of the form completely covering the engine was "entirely practical" under service conditions, but warned that "it must be carefully designed to cool properly."14 The NACA then announced to the press that aircraft manufacturers could install the low-drag cowling as an airplane's standard equipment for about $25 and that the possible annual savings from industry's use of the invention was in excess of $5 million- more than the total of all NACA appropriations through 1928.15
With the initial round of wind tunnel investigations completed, Langley borrowed a Curtiss Hawk AT-5A !airplane from the Army Air Service, fitted it with the J-5 engine, and applied cowling no. 10 for flight research. The Hawk's speed increased from 118 to 137 miles per hour with the low-drag...
....cowling, an increase of 16 percent. The results of the instrumented flight tests had enough scatter for Langley to have been justified in claiming a 20-mile-per-hour speed increase instead of 19, but the NACA kept its advertised figure conservative.16
Effectiveness of the cowling was demonstrated to the public almost immediately. In February 1929 Frank Hawks, who was already famous for his barnstorming and stunt flying, established a new Los Angeles-to-New York nonstop record (18 hours, 13 minutes) flying a Lockheed Air Express equipped with an NACA cowl that inreased the aircraft's maximum speed from 157 to 177 miles per hour. The day after the feat, the Committee received the following telegram:
A few months later, the NACA won its first Collier Trophy, for the greatest achievement in American aviation in 1929. This pleasant recognition not only promoted the cowling's economic value and justified the NACA's decision to build the PRT; the award was also timely support for the NACA's request for money to build 41 full-scale tunnel.18
A second stage of systematic cowing research had begun in late 1928 - even before the public acclaim - and involved tests with several different forms of cowling, including individual fairings behind and individual hoods over protruding cylinders, and a smaller version of the new complete cowling, all mounted on an open-cockpit fuselage. The researchers at Langley also performed drag tests with a conventional engine nacelle and....
....with a nacelle having the new complete design. The individual fairings and hoods proved ineffective in reducing drag, and it was found that for a smaller body as opposed to a fuselage with larger cabin, the complete cowling reduced drag more than twice as well as the conventional cowling did. Data from the AT-5A flight tests confirmed this conclusion.19
In 1929 Langley mounted its low-drag cowling on the engines of a Fokker trimotor. When comparative speed trials proved extremely disappointing, the engineering staff started to wonder how the position of the nacelle with respect to the wing might affect drag. In the case of the Fokker (as well as the Ford) trimotor, the original design location of the wing engines was slightly below the surface of the wing. As the air flowed back between the wing and nacelle, and the distance between them increased toward the rear of the nacelle, the expansion required was too great for the air to flow over the contour smoothly. The PRT team tried fairing-in this space, but achieved only a small improvement.20
Nevertheless, the lab's systematic, empirical approach soon yielded its dividend. With the help of his assistants, Fred Weick laid out a series of model tests in the PRT with NACA-cowled nacelles placed in 21 different positions with respect to the wing-above it, below it, and within its leading edge. The resulting data on the effect of the nacelle on the lift, drag, and propulsive efficiency of the airplane m de it clear that the optimum location of the nacelle was directly in line with he wing, and with the propeller fairly well ahead. Although their primary emphasis was on drag and improved cooling, the tests at Langley also confirmed that a complete cowling of the radial engine, if situated in the optimum position, could in some cases actually increase the maximum-lift coefficient.21 The NACA transmitted....
.....these results confidentially to the army, navy, and industry. (This private transmission was very significant: it gave U.S. industry several months lead time over European aircraft builders.) After 1932 nearly all transport and bombing airplanes with radial, wing-mounted engines - including the DC-3, the B-17, and many other famous aircraft of the era that followed - used the NACA cowling and located the nacelles with reference to the NACA data.
The cowling was winning so much respect in the late 1920s and early 1930s that the NACA seemed to have gradually identified itself more and more with the systematic experimental approach that had been the basis of that successful research. In 1936 the head of the Aerodynamics Division, Elton W. Miller, reported to engineer-in-charge Henry Reid that "an effort is being made throughout the Laboratory to conduct every investigation in as thorough and systematic a manner" as the cowling program.22 The following year, George Lewis in Washington told Reid to frame and hang in his office or along the corridor of the LMAL administration building a copy of a quotation from a recent speech by President Hoover in praise of Thomas Edison:
Clearly the pattern of work behind the cowling-the NACA's greatest public success to date-was contributing to a clearer sense of institutional identity and mission.
At least one contemporary observer saw this identification with systematic engineering as unflattering to Langley laboratory. Frank Tichenor, the outspoken editor of the journal Aero Digest who had hired Max Munk, labeled the NACA cowling "a development rather than an original work" and misjudged it as being far less effective than the Townend ring, a rival concept developed simultaneously by Hubert C. Townend at the British National Physical Laboratory.24 Though the NACA can perhaps be criticized  for trying to take too much credit for industry's adoption of the cowling, one must underscore the truth that the NACA never really claimed to have invented the cowling. It professed neither conceptual originality nor revolutionary development. What the NACA did claim-and what seems beyond dispute-is that the PRT permitted engineers to work with full-scale cowled engines. Better experimental equipment had led to more comprehensive and more useful data. It is not so clear in retrospect, however, whether the NACA's commitment to the pattern of experimental parameter variation for the next stage of cowling research signified technological momentum, or technological inertia.
The third stage of cowling research, 1931-1934, began at Langley when many more aircraft manufacturers decided to adopt the NACA design as standard high-performance equipment. A few companies did rather well with their applications of the NACA no. 10 cowling, especially those that put a series of adjustable flaps around the circumference of the metal jacket in the hope of better regulating the release of used air. (Those that tried to encourage more cooling flow by employing larger exit openings failed, however, sometimes to the point o nullifying the external drag advantage.) With the development of twin-row engines such as the Pratt and Whitney R-1830 of 1933 and 1934-with one row of cylinders behind the other whole new problems arose.25 This situation challenged Langley to obtain more trustworthy data on the general aerodynamic properties of the proven NACA design. Practical results had been obtained from experimental parameter variation, and they had seen used profitably. Now it was time for a clearer understanding of them, s that still more results could eventually be achieved.
Three major branches of the laboratory became involved in the ambitious program. The Power Plants Division worked to improve the efficiency of radial engine cooling by varying such engine parameters as pitch, width, thickness, and shape of the fins. The 7 x 10-Foot Wind Tunnel section, using small models, sought the best possible cowling arrangement for necessary cooling with minimum drag by streamlining the front and rear openings, changing the size of the nacelle, and altering the camber of the cowling's leading edge. The PRT team was then to verify the results of the tests made by the other two groups. Full-scale propeller-cowling-nacelle units were to be tested under conditions of taxiing, takeoff, and level flight.26
Though the first two parts or the program advanced without much difficulty, the PRT tests-the final and most important part-ran into major problems soon after starting in 1933: the 100-mile-per-hour tunnel could simulate only the climb speeds of the cowled engine being used (a borrowed Pratt and Whitney Wasp); the obsolete shell-type baffles employed to  deflect cooling air toward the hottest parts of the engine were too loose for the NACA researchers to work with effectively;27 and, more importantly, certain anomalies that no one at the lab could explain plagued the cowling drag measurements. Together these problems contributed to a growing "maze of contradictory data" about cowlings. Despite five years of NACA experimentation and three years of general industrial flight test experience, American aeronautical engineers felt a "general suspicion" that there was "something mysterious or unpredictable determining the efficiency of engine."28
To move beyond the paralyzing confusion of this experimental impasse, Langley's cowling research needed some analytical help. It was eventually provided by the head of the laboratory's small Physical Research Division, Theodore Theodorsen. A Norwegian-born engineer-physicist with a trigger mind and tremendous power of concentration, Theodorsen had already seen in Langley's pattern of airfoil testing in the VDT the need for experimental routine to be fertilized with a stronger dose of theory (as the terms of his opposition to Eastman Jacobs's idea for a new low-turbulence VDT, outlined in the previous chapter, plainly showed). In the curious introduction to his seminal 1931 report on the "Theory of Wing Sections of Arbitrary Shape"-curious at least in an NACA report for stating a bold personal opinion and implicitly taking part of the parent organization to task-Theodorsen had asserted that
What Theodorsen believed the NACA needed in order for it to move beyond the impasse temporarily blocking the progress of its experimental cowling program was more attention to the "pencil-and-paper" work that could lead to a complete mathematical and physical understanding of the basic internal and external aerodynamics of the different cowling shapes.29 And what this meant in terms of the history of Langley's method of cowling research was....
....a turning away from experimental parameter variation, and toward that distant goal of complex understanding.
Theodorsen first perceived new cream to be skimmed off the top of the old cowling and cooling investigation while serving on the editorial committee that reviewed the draft report on the tests of the full-scale propeller-cowling-nacelle units in the PRT. After pointing to the blunt afterbody of the nacelle as the probable source of the anomalies that had been observed in the drag data, he suggested to his colleagues that the stalled cowling program could be completed as planned (and his resolution of the drag anomalies verified) by a new, more comprehensive and analytical full-scale investigation. Its aim, underscored Theodorsen, would be both to improve basic understanding of the obscure cooling mechanisms of the cowled engine and to put the understanding of the relationship between internal flow and drag on a more rational basis. The provocative suggestion was adopted; the engineer-in-charge transferred most of the cowling work and some of its key workers to Theodorsen's division.30
Previously the PRT research team had focused almost entirely on the net or overall effect of the cowling on drag and engine temperatures. What Theodorsen now proposed was to investigate the fundamental flow involved. In part, the approach of Theodorsen's new cowling research team still followed that of experimental variation. The Wasp engine having proved  inadequate as part of the test bed, they built a full-scale wind tunnel model with a dummy engine, which had one cylinder heated electrically. Numerous combinations of more than a dozen nose shapes, about a dozen skirts, six propellers, two sizes of nacelles, and various spinners were tested. But hoping to produce a detailed handbook by which designers could better understand the actual functioning of the NACA cowl, they also included extensive measurements of pressure in both the external and internal flows.
Langley's revised cowling program thus remained primarily experimental, but it now also allowed quantitative analysis and computation of these flow pressures. This quantitative analysis, which had been lacking in the previous work, eventually produced some new NACA cowling designs, but more importantly it provided solid answers to virtually all the remaining questions about the fundamental principles of the cowling and cooling of radial engines.31 It demonstrated conclusively that the early NACA designs had been "quite haphazard and often aerodynamically poor" and had cooled the engine successfully only by a crude excess of internal flow and internal drag (a conclusion that Vought engineers had apparently arrived at on their own, earlier, on behalf of Pratt and Whitney and its R-1830 engine).32 Designers of future cowlings, like airfoil designers, would have to be much more sensitive to such subtleties as the ideal angle of the cowling's leading edge attack on the local airflow. The work even demonstrated as fact something that everyone had unconsciously assumed to be physically impossible when the cowling research began in 1926: a proper engine cowling could, by making the enclosed baffled engine act in essence as a ducted radiator for cooling, lower operating temperatures more than could full exposure of cylinders in the airstream. With an understanding at once basic and advanced, the national aeronautical establishment could now begin to focus on more specific, higher-speed applications of cowlings, work that would be essential to the design of military aircraft used by America and her allies during World War II.
The history of the NACA's cowling research from 1926 to 1936 celebrates a victory but also demonstrates an important general point about research: No matter how practical or otherwise advantageous any one method may be, it always has some disadvantages. Systematic parameter variation had enabled the researchers at Langley to delineate a cowling that significantly reduced the drag of a radial engine without degrading its cooling, but because initial success came rather quickly and easily, they did not have to understand exactly why the cowling worked. When questions and doubts....
....arose, and data seemed contradictory and mysterious, the original empirical method was unable to proceed. Only then did Theodorsen design the research program whose goal was an understanding that went far beyond the mere collection of overall performance data on a variety of promising but arbitrary shapes. The cowlings that resulted from the Theodorsen program did not beat the earlier shapes as regards external drag (which is only a weak function of cowl shape), but with the tight baffles, small exit areas, and low internal drag made possible by the NACA's new criteria of understanding, the total drag of Theodorsen's shapes was dramatically less.
 Historians have tended to treat the NACA cowling as a magical piece of tin wrapped around an engine and producing fantastic results. As a result, they have failed not only to appreciate the systematic character of the laboratory work which made the initial design breakthrough possible, but also to pick up on the later work by Vought and Theodorsen which made the important breakthrough in understanding possible. The success of the cowling was not due to magic. Nor was it the result of simple cut-and-try or advanced theory demonstrating its ultimate superiority over empiricism. Rather, the cowling was the product of fruitful engineering science: a solid combination of physical understanding, intuition, systematic experimentation, and applied mathematics.
Ultimate success in research is never inevitable, however. Without the help of Theodorsen or someone else with comparable analytical and mathematical talents, cowling research at Langley might have remained indefinitely at the point of impasse.
* A version of this chapter appeared in the Fall 1985 issue of Aerospace Historian.