2.1 Transonic truss-braced wing (TTBW)

The earliest aircraft designers recognised the benefits of structural bracing as a means to enable the creation of thin, aerodynamically efficient wings. It was not long, however, before the cruise speeds of aircraft increased to the point where the aerodynamic performance penalties of external bracing (i.e. drag) began to outweigh the structural efficiency benefits that bracing could provide in all but lower speed propeller driven aircraft. As a result, there are many examples of low-speed aircraft in service today that employ the use of wing-supporting struts or wires. Fortunately, advanced computational tools are increasingly available to enable the study of Transonic Truss-Braced Wing (TTBW) technology as applied to a modern, transonic aircraft as shown in Fig. 2.

Figure 2.

Transonic Truss-Braced Wing high/low-speed and aeroelastic wind tunnel models [Reference Harrison, Beyar, Dickey, Hoffman, Gatlin and Viken6].

The significant potential improvements to aircraft aerodynamic efficiency enabled by ultra-high-aspect-ratio wings is an important incentive for further development of a TTBW. In 1977 Werner Pfenninger postulated that the use of truss-bracing could enable ultra-high-wingspan designs with extensive laminar flow [Reference Pfenninger4]. In theory, these designs could achieve lift-to-drag ratios approaching 50. These theories were revisited in 1999 by a team of researchers at Virginia Tech, who performed MultiDisciplinary Design Optimisation of a transonic commercial transport with a strut-braced wing [Reference Gern, Gundlach, Ko, Naghshineh-Pour, Sulaeman, Tetrault, Grossman, Kapania, Mason and Schetz5]. Work performed at Virginia Tech (and other academic institutions around the world) has continued this research, producing articles and papers too numerous to list.

In 2008, building on early collaborative work with Virginia Tech, Boeing initiated the Subsonic Ultra Green Aircraft Research (SUGAR) program to explore advanced aircraft concepts with the potential for dramatic reduction in fuel burn, noise and emissions targeting a 2030–2035 entry into service. Over the past 12 years Boeing has worked with NASA under the Subsonic Fixed Wing (SFW) program to study a variety of different aircraft configurations and technologies. These studies have in recent years focused on understanding the promising potential of the TTBW concept as applied to a modern commercial transport. While the original SUGAR TTBW aircraft concept flew at Mach 0.7, these studies have evolved to focus on a Mach = 0.80 design. Computational studies and wind tunnel test programs under SUGAR have shown a potential fuel burn reduction of ∼9% for a 3500nm mission as compared to a cantilever-wing design of equivalent technology [Reference Harrison, Beyar, Dickey, Hoffman, Gatlin and Viken6]. The objective of the latest phases of the SUGAR program has therefore focused on further investigating and maturing the TTBW concept within the context of its potential applicability to modern commercial airliners. The primary attributes of the TTBW configuration that enable its potential to achieve these gains are:

• a dramatic increase in wing aspect ratio, resulting in decreased lift-induced drag

• structural efficiency from strut-bracing, resulting in decreased wing root bending moments

• high-wing engine installation, enabling efficient integration of large-diameter high-efficiency propulsors

These design features have promising prospects for improved vehicle performance; however, they are not easily realised. The integration challenges of these same novel features result in design considerations not typically examined on a typical cantilevered-low-wing design. A (non-prioritised) subset of some of the primary considerations are:

• non-linear structural design

• non-linear aeroelastic behaviour

• highly coupled transonic lifting surfaces

• high-speed buffet

• thin-wing actuation systems

• high-lift system design and integration

• thin-wing icing

• achievable fuel volume

Over five phases, the SUGAR program has worked to systematically address these challenges through a series of design and testing activities. The design activities included structural design and analysis, detailed aerodynamic design for cruise Mach numbers ranging from M = 0.7 to 0.8, aeroelastic design and analysis, and high-lift system design. To date, the program has also completed six wind tunnel tests of the configuration to validate transonic cruise performance (in the NASA Ames 11-foot Transonic Wind Tunnel), high-lift performance including in-ground-effect and icing (in the NASA Langley 14- by 22-Foot Subsonic Tunnel), and aeroelastic behaviour (in the NASA Langley Transonic Dynamics Tunnel). Future investigations have also been proposed to increase understanding of additional aspects of the vehicle design and integration, including additional high-lift testing, acoustics quantification, more detailed icing investigations, and structural design and testing. NASA is also investigating the possibility of a large-scale flight demonstrator of the TTBW concept, which would continue the TTBW design validation efforts.

2.2 Flying V-configuration

The Flying V configuration is a radically new configuration for a long-haul passenger aircraft. The passengers, cargo and fuel are located in the wing (Fig. 3). It is estimated to have 20% higher payload-range efficiency than its tube-and-wing counterpart for the same top-level aircraft requirements [Reference Oosterom7]. This is caused by three factors. First, the absence of a distinct fuselage and tail reduce the wetted area by 5% leading to reduced friction drag. Second, the large winglets increase the effective span of the wing leading to a reduction in lift-induced drag. Finally, the lateral distribution of the payload and fuel reduces the bending moments and thereby the structural weight of the aircraft. These benefits stem directly from the shape of the aircraft and can be further complemented by innovations in the airframe or the propulsion system. Research into the viability of the Flying V is currently being performed primarily by researchers at Delft University of Technology. In the subsequent paragraphs, the Flying V design is further described and the associated challenges with this new configuration are discussed.

Figure 3.

Planform view of the largest version of the Flying V configuration with 52 business class seats (55” pitch) and 309 economy class seats (32” pitch) [Reference Oosterom7].

As can be seen in Fig. 3, the cabin of the aircraft features a 3-4-3 configuration in both legs yielding four aisles. This reduces the boarding and deboarding time [Reference Isgro8]. A large cross aisle is present that connects the legs of the cabin with a large galley located in the middle. The door distribution is chosen such that the emergency evacuation can be ensured within 90 seconds for a high-density configuration of the cabin. The cargo compartment is located behind the cabin. To allow for the cargo nets to expand during a forward deceleration of 9g, space is reserved between the cargo containers and the aft cabin wall. The cargo containers that are used are of the LD4 type and are located in a tapering part of the wing. The fuel tank is located behind the cabin in the trailing edge of the wing as well as in the wing box of the outer wing. The economy seats are rotated by 9 degrees with respect to the cabin aisles to stay within certification constraints. The business class seats have seat belts with embedded airbags. By removing wing plugs (denoted in green and yellow in Fig. 3), smaller versions can be created such that a family of Flying V aircraft results.

The Flying V has a crescent wing planform with a 64-degree leading-edge sweep on the inner wing and a 40-degree sweep on the leading edge of the outer wing. The large inboard sweep angle allows for a cabin height of 2.1 meters without the formation of strong shock waves in the design condition [Reference Faggiano, Vos, Baan and van Dijk9]. The outer wing is designed as a traditional transport wing with supercritical wing sections. The low aspect ratio (around 4.5) combined with the high sweep angle result in a low lift-curve slope. Due to the large sweep angle of the inner wing, the stall of the Flying V is quite gradual with a maximum lift coefficient of around 1 [Reference Palermo and Vos10]. Depending on the Reynolds number, large vortex structures are formed beyond a critical angle of attack (see Fig. 4) that result in vortex lift. Since the outboard wing stalls first, there is a risk of a strong nose-up pitching moment. Therefore, a leading-edge slat or droop nose might be necessary to ensure acceptable low-speed stall characteristics.

Figure 4.

Large vortex structures appear at high angles of attack over the inner wing of the Flying V [Reference Viet11].

The outer wing features various types of control surfaces. The most outboard control surfaces on the wing are multifunctional. They act as drag rudders and elevons. Together with the “normal” rudders on the large winglets, they provide sufficient yawing moment to control the aircraft in a one-engine-inoperative condition [Reference Cappuyns12]. The inner two control surfaces act as elevons controlling both pitch and roll. Secondary flight control surfaces in the form of spoilers are added on the inboard and outboard side of the engine. They ensure that 10% of the lift can be dumped upon landing. To ensure adequate flying and handling characteristics the aircraft is statically stable in pitch, roll and yaw. With the addition of a simple yaw damper and auto-throttle also dynamic stability is ensured in all modes (see Fig. 5). However, due to the absence of a tail, a strong coupling between the pitching and heaving movement is experienced which warrants the introduction of dedicated flight control laws for this configuration [Reference Garcia, Vos and de Visser13].

Figure 5.

Flight control was experimentally verified using a 5%-scale flight-test demonstrator. Photo: Joep van Oppen for KLM and TU Delft.

The inner wing of the Flying V is pressurised. Therefore, an oval cross section is used with a piecewise circular perimeter. To ensure the structure does not deform upon pressurisation, a trapezoidal box structure is added inside this oval shape. The lower member of this trapezoid acts as the passenger floor. The outer wing primary structure is conventional, consisting of a box structure. The lateral distribution of fuel and payload reduces the stresses inside the structure compared to a conventional wing. This effect has been shown to outweigh the increase in stress due to the noncircular shape of the cabin cross section [Reference Claeys14,Reference Van der Schaft15]. However, further work is needed to detail the structure and assess it on a variety of loads such as crash loads or ditching loads. While the flight performance during climb, cruise and descent are not notably different from a tube-and-wing aircraft, the take-off and landing characteristics are quite different. Due to the low lift-curve slope of the wing, the aircraft has a maximal approach attitude of 14 degrees. Note that the aircraft does not feature any trailing-edge flaps to reduce its landing attitude. To allow the aircraft to have sufficient clearance at the tip during an 8-degree banked landing, long landing-gear legs are required and perhaps even an outrigger wheel (see Fig. 3) [Reference Bourget16]. However, operational constraints place the maximum height of the cargo door at 5.5 meters when the aircraft is completely empty. To stow the main landing gear efficiently in the wing, a double hinged mechanism is required [Reference Bourget16]. Furthermore, for a maximal approach speed of 140kts at the maximum landing weight, an overnose angle of 31 degrees is required to satisfy certification constraints on pilot visibility [Reference Van der Pluijm17]. This drives the shape of the cockpit, which is further complicated by the stowage of the nose wheel combined with the relatively low height of the fuselage.

The over-the-wing engine poses challenges in terms of structural integration, aerodynamics, noise, and maintainability. It is expected that the Flying V will utilise an ultra-high bypass-ratio engine, although it could also be suitable for an unducted fan. Due to the absence of wing reflection and the partial shielding of the inlet, the Flying V engine causes less community noise than an under-the-wing engine. Research is being conducted to investigate two integration options: positioning engine and nacelle on a separate pylon or blending the nacelle with the wing and allowing the wing boundary layer to be ingested by the inlet. The latter results in less wetted area, lower structural weight and easier maintenance while leading to lower total pressure recovery and higher fan distortion.

Research on the Flying V is still relatively recent, and it is still to be confirmed whether this configuration can meet all the certification requirements, while still achieving the improvement in payload-range efficiency.

2.3 Electrification of aviation

Aviation is undergoing a renaissance ushered in by a convergence of technologies that is enabling new aircraft designs, new missions, and new aviation business models. One of the most promising of these technologies is electric propulsion, which is fostering the greatest change to the design of aircraft since the advent of the jet engine. Electric propulsion comprises the reinvention of the aircraft powertrain, replacing fuel-burning internal combustion engines with electric motors, and replacing aviation fuels with energy stored in battery cells. Electric propulsion is now viewed as increasingly viable for aviation largely because of the recent progress in battery technology, power electronics, and motor technologies which have been advanced by industries ranging from consumer electronics to power tools. Underpinning these advances have been an array of fundamental breakthroughs stemming from fields as diverse as electrochemistry, material science, solid mechanics, microelectronics, and control theory.

Electric propulsion should not be thought of as a one-to-one replacement for fuel burning propulsion systems due to the relationship between the mission and requirements for an aircraft and the technology available for use in the aircraft’s design. In particular, the capabilities of any technology enable certain missions and requirements to be met practicably and affordably, and the limitations of the technology prevent other missions and requirements from being met practicably and affordably. For example, the unique capabilities of electric propulsion have created new categories of small UAVs, both fixed wing and VTOL. These new aircraft concepts are enabling new missions, allowing flight to be used in new and innovative ways.

All technologies have strengths and weaknesses, requiring decisions by aircraft designers to select the best technologies or combinations thereof to meet any particular mission and requirements. In this context, electric propulsion simply represents another choice of aircraft propulsion technology to add to existing options. For certain missions and requirements, electric propulsion will be feasible, and for other missions and requirements, conventional fuel burning propulsion will be feasible. Indeed, because electric propulsion is in its technological infancy, the space of possible missions and requirements that it can satisfy is smaller than that of conventional propulsion, most notably in terms of range performance. For a subset of missions and requirements, both electric and conventional fuel burning propulsion are technically feasible, and trade-off studies will determine which solution is preferred. Figure 6 illustrates these concepts of how electric and conventional propulsion address different missions and requirements.

Figure 6.

Electric, hybrid electric and liquid fueled propulsion in the context of aircraft missions.

Considering these trade-offs, we can expect that aircraft designed for some missions and requirements may transition to electric propulsion, while other designs retain conventional fuel burning propulsion systems including piston engines and gas turbines. Other missions and requirements may motivate hybrid-electric propulsion solutions that leverage both fuel burning engines and electric propulsion components. The mission space depicted in Fig. 6 is not static, as technologies advance, we can expect the feasible mission bubbles for each category to grow.

That not all aviation missions and future aircraft designs will transition to electric propulsion does not diminish the importance of this emerging technology but rather frames its true opportunities. These opportunities lie in the space of requirements in Fig. 6 that electric propulsion can address but that conventional fuel burning propulsion cannot. This space includes requirements for emissions-free flight, low noise, and low maintenance costs. By enabling new aircraft designs that can address these requirements practically and economically, electric propulsion will enable aircraft to fulfill missions entirely new and transformational to aviation.

While the electrification of aviation represents an exciting future, we must anchor our perspective on the future on what has already been achieved. Figure 7 depicts a mosaic of electric and hybrid-electric aircraft; some piloted and some not, some production, representing hundreds of thousands of units sold, and some prototype. Some flight missions are already dominated by electric solutions — as electric component technologies move forward, the missions accomplished using electric propulsion will only expand.

Figure 7.

Contemporary examples of battery electric, solar electric and hybrid electric aircraft.

3.1 Empirical methods and data

Although future aircraft design process developments will someday combine model-based design (MBD) and artificial intelligence (AI) and fundamentally change how aircraft are designed, for now and the foreseeable future empirical design methods and data continue to start and sometimes complete the aircraft design process. The rationale is clearest for design databases since AI tools will almost certainly be database driven. The issue for empirical methods, however, is not whether they will continue to be viable and efficient which has already been established by Brandt and others [Reference Brandt44]. Rather, the issue is whether they will continue to (1) be applied in the design room and (2) taught in the classroom where future designers learn design fundamentals. The answer to the former will almost certainly be yes as discussed in section 3.2. However, the empirical lineage may not be obvious to users as the methods become increasingly integrated into higher-order codes as discussed in [Reference Carty45]. The most concerning issue is whether empirical methods continue to be taught as discussed in section 3.2.

From a historical perspective, empirical/data-based design methods trace to the 1930s when Professor K. D. Wood of Cornell University published the first of his Airplane Design series which was data driven and iterative. It started with requirements and an initial concept followed by data-based design cycles at increasing levels of fidelity intended to converge to a realistically sized concept capable of meeting requirements. The method used data from prior aircraft and essentially is what we know today as an empirical design method.

In its review of Wood’s text in 1935, the RAE Journal compared Wood’s approach to that used in British universities which we paraphrase as “consisting mainly of theoretical instruction with much of the information in the book not becoming known to most until arriving in industry.” Apparently, the reception to a text was sufficient for Prof. Wood to publish 10 editions through 1963 followed by his better-known Aerospace Vehicle Design Volume I – Aircraft Design [Reference Wood46]. The 1963 text devoted 189 pages to aircraft and rotorcraft design methods (mostly empirical) and another 238 pages to supporting design data.

Later followed Professor Roskam’s well-known, eight (8) volume Airplane Design series first published in 1985 [Reference Roskam47]. Roskam took empirical design methods to a higher-level of technical development. So, today aircraft design engineers and instructors have access to a wide range of excellent texts that combine empirical and physics-based methods across the aircraft design spectrum including books by AIAA Aircraft Design Technical Committee authors [Reference Raymer48–Reference Gundlach52].

One question for consideration here is where are empirical design methods and data headed or more importantly, how do they need to change to meet future needs as discussed in section 3.2? The first is an easy answer. Based on current trends, all methods and data will get absorbed by higher level master codes and at a detail level, may be known only to those involved in the care and feeding of the code. Working level engineers will give the codes problems to solve and let it pull up whatever it or engineering management considers the most applicable method. Working level engineers will still need to be familiar with the fundamentals of these methods so it’s important that they stay in the curriculum, or they really will get lost. The answer to the second question is that there are important things that need to be done today to make sure the most important methods and especially the data stay relevant for the next century of aeronautics.

3.1.1 Aircraft empirical methods and data: Industry design applications

This is a broad subject with many examples, most of which are proprietary. For established aircraft companies, they and their predecessors’ companies have decades of experience and data widely applicable across existing product lines. Methods are verified and often certified against the company database which makes the two essentially inseparable. In concept most company applications are similar and deal with a broad range of product design, analysis, test and support issues.

When companies are new and/or moving into new product areas, their database will be thin and the method applications risk will go up, but their basic technical approach typically doesn’t change. The process can start with a clean sheet or an existing product or some place in between. Regardless there is a need for quick initial answers with enough fidelity to make good decisions about whether to proceed. And this is where empirical methods are most widely applied. An example is the Roskam method shown in Fig. 8.

Figure 8.

A typical conceptual aircraft design methodology [Reference Roskam47].

The Roskam method consists of two series of sequential conceptual design loops called Class I and Class II with the difference being fidelity. In both classes, empirical and physics-based methods are used with most Class I being empirical. For example, initial mass property estimates are almost always empirical with the difference between classes being fidelity. For example, a Class I empty weight estimate might be based on an empirical gross mass data as shown on the left side of Fig. 8 or built up by component such as the wing mass based exposed wing area and type shown on the right side [Reference Roskam47]. Class II might break the wing down into the wing box and control surfaces and flaps. Class I empennage sizing is also almost always initially based on historical tail volume coefficients which go through critical sizing checks for nose gear liftoff in Class II. Early aerodynamics and volume required estimates are also empirical. Class II aero estimates are typically physics based (CFD) but typically address major components with details such as intersections and fairings estimated from data. Structural layouts may not be called empirical, but they are based on prior vehicle layout experiences as are propulsion installation loss estimates. An example of a physics-based Class I is a static balance calculation based on estimated component mass and location, some of which such as alighting gear location are based on experience. Regardless of the specific methods, the objective is almost always the same – assess the concept and see if it works. Usually, the first one does not so all early cycles need to be done quickly [Reference Chaput and Mark53]. In fact, the motto of my design room at General Dynamics Fort Worth was “you have to kiss a lot of toads before you find a prince.” [Chaput]

Even after a concept is evaluated and selected for development, empirical estimates continue in use. For example, at a preliminary Design Review where the objective is to freeze the outer mould line and major structural locations, component masse beyond the “big bones” are estimates as are target weights [Reference Chaput54]. Even at a critical design review (CDR) where the objective is to get permission to start releasing drawings to the factory, most smaller airframe component masses continue to be empirically based. So, the prospects for continued application of empirical methods in industry are almost guaranteed and the only place where they will continue to be called into question are at places that do not do real designs, i.e. at universities.

3.1.2 Aircraft empirical methods and data: Design education applications

Although today’s texts apply a variety of methods, the fundamental concepts are similar in that the initial objective is to educate students on how to define a realistically sized starting concept that should work. A second common attribute is the use of data to ensure that intermediate and final results are realistic. What is typically not emphasised is industry-standard, multi-concept exploration and iteration towards the objective of identifying a “best” concept vs. one that simply works. The reason is design course schedule constraint, even empirical concept exploration methods are time-consuming. One solution is parametric model-based design (PMBD) education where traditional empirical methods are integrated with physics-based parametric geometry models [Reference Chaput and Mark53]. Once developed by a student, a PMBD model can be used to calculate performance and also morphed into a variety of alternative concepts for comparison and ‘optimisation’ to best meet design requirements. The term ‘optimised’ is widely used in the classroom but often misunderstood by students. Optimisation is process of multi-discipline compromise to achieve a compliant but better-balanced design.

Regardless of design methods used in the classroom, a common issue for all instructors is student understanding of how the methods are applied which, by definition, involves comparisons to design data. While the future could bring design education processes that enable theory to be applied without comparisons to real aircraft, one must question the educational rationale. On the job, students will be dealing with real products that have to work across multiple users and disciplines, so taking real products out of the design education process would be a return to the pre-1935 methods discussed in the introduction. The key issue, therefore, is the availability of data, its form and how it is applied in design education.

The availability issue has been with us for a long time and is getting worse. Almost everything about a commercial aircraft design below the empty weight (EW) level is considered proprietary, particularly mass properties. The situation for military aircraft is similar except the issues are not only proprietary but national security related. Yet some of the most recent data in education design databases is for military aircraft while commercial aircraft of similar vintage are missing. In fact, about the only detailed mass property data on civilian aircraft is for military variants of the same aircraft, which is indeed strange.

The form of the data is another issue. Much of our aeronautical data goes back to the period from 1930 through 1960 where data was plotted graphically as a part of the analysis process. The plot variables were traditional and typically two-dimensional. Much of the aerodynamic and propeller data is still relevant and widely used. The basic form, however, is graphical which limits application of modern, multi-dimensional analysis methods. What’s needed is digital data and it’s not widely available.

Another issue is availability of data on modern aircraft or modern forms. Except for companies trying to break into a new product line, this may not be an issue for industrial designers where the most important data is in their own product archives because they reflect internal processes, procedures and capabilities. But for educational purposes our data base is not only old and cumbersome to analyse but so old as to be a turn-off for students, including for research.

If the data availability and form issues can be mitigated as discussed in Section 4, the next most important issue is how we teach student to think about and use design data. It is a fact of life that engineering education emphasises equations, especially derivations. And one design education issue is equations derived from data. An example is a widely known data plot used for initial aircraft sizing in the form of a log plot of flight design gross weight (FDGW) vs. EW as shown on the left side of Fig. 9.

Figure 9.

Comparison of curve fit vs. physics approaches to design data.

FDGW is approximately the sum of EW plus flight design useful load (UL) where UL is driven by requirements that can vary widely, even within aircraft types. Nonetheless, students learn to make initial estimates of EW from FDGW using curve fit equations with emphasis on the equation. What is important, however, is not the equation but the physics. For example, plotting the same data in a wing loading form as shown on the right side of the figure and discussing the relationship with lift coefficient (C_L), teaches the concept of physics-based limits within which requirements and designer choices define the design outcome.

Once students learn to start thinking about the physics of design data, they are better prepared to intellectually deal with our most widely misused data-based relationships in aerospace, which are mass property parametrics where curve fit equations are all students see. The equations mask the underlying data scatter and constraints and leads to over-reliance on equations vs. data. As a result, students routinely apply the relationships where they do not apply and are confused about why things did not work out in the end.

3.2 Future trends

Since the earliest days of aircraft design, historical data and derived empirical relationships have been the basis for sizing and performance correlation. The methods are still widely used and applicable and will probably stay that way for the foreseeable future. However, as modern design tools and methods continue development, the underlying data and relationships, are likely to become imbedded in higher level methods and lose their identity and understanding including limitations on application. It is important, therefore, that the basic methods be taught, understood and properly applied by future designers.

The antiquity of current publicly available design data is a serious issue for design education, especially for commercial aircraft. Contemporary data plots are often limited to first- and second-generation jet aircraft that have been out of service for longer than most students have been alive. The issue is the proprietary concerns of manufacturer legal departments that apparently do not recognise the silliness of maintaining close hold of antiquated data. It helps fuel the concept among students that aeronautics is old fashioned when in fact behind the proprietary walls, aeronautics is as cutting edge as any other discipline.

Although it is unlikely that manufacturers will yield on proprietary concerns, they should at least be willing to release data to trusted agents able to sanitise data to mitigate proprietary concerns that could give the design database a significant refresh. A useful but dated model of the trusted-agent concept was when the U.S. Air Force, which tasked the RAND corporation to sanitise aircraft manufacturer cost data and generate cost estimating relationships still used today, albeit based on ancient data that also needs a refresh [Reference Boren55].

Another option is for sponsored research where well-documented historical aircraft such as early jet transports are updated with modern materials, subsystems and propulsion. Included would be validating the results against a sanitised modern counterpart aircraft. These would not only be good graduate student team research projects but help teach modern methods of design.

A final issue is the 2D data plot format of design data which is still highly relevant. Those plots may have been useful for yesterday’s engineering, but today’s data needs are digital. This means funding is needed to republish prior reports with digitised data files.

3.3 Multidisciplinary Design Optimisation (MDO)

Aircraft design requires a team of engineers and leaders transform a set of requirements into mathematical models, then into blueprint drawings and specifications, and finally, into physical hardware. Aircraft design requires expertise in diverse engineering specialties including applied mathematics (computer programming and numerical analysis), aerodynamics (for external and internal flows), acoustics, thermodynamics, materials science, chemical engineering, electrical engineering (power generation, RF applications and control theory), structural mechanics (statics and dynamics), machine and mechanism design as well as manufacturing engineering and operations research. Aircraft design also requires engineers to interact with a broad set of non-engineering disciplines, including law, finance, marketing, accounting, meteorology, supply chain, industrial design and even diverse professions such as fashion merchandising [Reference Takahashi56].

Aircraft sizing has always been a numerical, multi-disciplinary exercise. Traditionally, this work was distributed amongst many engineers working in relatively siloed disciplines; refer to Fig. 10. Classical aircraft sizing depends upon the use of highly simplified models that capture the multi-disciplinary interactions among weights, aerodynamics, propulsion and performance [Reference Hays58]; this is expressed in Loftin’s thrust-loading/wing-loading constraint chart [Reference Loftin59] made famous by Roskam [Reference Roskam47] amongst others (Fig. 11).

Figure 10.

Classical aircraft design problem [Reference Carty45].

Figure 11.

Classical thrust loading/wing loading constraint chart.

Multi-disciplinary optimization (MDO) grew from operations research. In many ways, the former U.S. Defence Secretary, Robert McNamara, is the father of MDO. He and his other ‘whiz kids’ got their start during WWII tasked with providing a scientific method to provide a quantitative basis for decisions regarding the operations under their control [Reference Morris, Williams, Ahlberg, Chappell, McNamara and Glass60,61]. These defence studies performed critical analyses to aid the war effort. For example, to optimise the size of transport ship convoys as to best minimise losses. Operations research provided counterintuitive insight showing that vulnerability was most correlated to the number of escort vessels present, rather than the size of the convoy [Reference Gannon62].

By the end of the 1950s, operations research principles began to influence many attributes of design. Beginning with his 1960 landmark paper [Reference Schmit63], Professor Lucien Schmit lead the transformation of engineering design from an analytical (calculus and asymptotic expansion-based methods) pursuit to one involving direct numerical solutions. Schmit began his career at Grumman and later transitioned to academia. He and his student Garret Vanderplaats played a transformative role in introducing gradient-based optimisation methods to transform finite-element structural analysis methods into structural design synthesis codes [Reference Vanderplaats64].

After achieving considerable success with structural design and NASTRAN during the U.S. space programme, operations research inspired multi-disciplinary optimisation techniques grew in scope. During the 1980s and 1990s at NASA/Langley Research Center, Jaroslaw Sobieski championed decomposition methods specifically designed for MDO applications. Sobieski emerged from the NASA structures community [Reference Tolson and Sobieszczanski-Sobieski65] during the formative years of MDO, he collaborated with many academic groups, particularly those at Virginia Tech [Reference Grossman, Haftka, Kao, Polen, Rais-Rohani and Sobieszczanski-Sobieski66]. At the United States Air Force, Dr. V.B. Venkayya played a similar role; coming from a flutter and structural-dynamics perspective; he was instrumental in developing the ASTROS aero-structural design system [Reference Herendeen, Hoesly, Johnson and Venkayya67].

Modern MDO methods are composed of a variety of approaches. Many consider Isaac Newton as the grandfather of optimisation, having devised an efficient root-finding algorithm in 1669. This ‘gradient method’, which can be extended to consider millions of design variables [Reference Martins68], includes various methods including those which directly and indirectly (i.e. penalty functions) impose constraints as well as algorithms such as methods of feasible directions, conjugate gradients as well as sequential linear programming and sequential quadratic programming methods [Reference Vanderplaats69].

Alternatively, non-gradient-based methods, including genetic algorithms, simulated annealing and ant colony algorithms, have also become popular to apply to complex problems that do not present objective functions with clear local maxima [Reference Vanderplaats69].

Finally, modern computers with high-pixel density displays enabled the growth of design space visualisation methods. These tools directly involve the human to determine the preferred solution; they have become popular with complex multi-objective design problems with poorly understood constraints; see Fig. 12.

Figure 12.

System level sensitivities to design variables for a notional HSCT [Reference Mavris, Bandte and DeLaurentis70].

Commercial aircraft design software embodying MDO principles includes monolithic packages (which include all germane analysis and optimisation methods) and extensible frameworks. Daniel Raymer’s RDS [Reference Raymer71], NASA’s FLOPS [Reference McCullers72] and Professor Jan Roskam’s DAR Corporation AAA [Reference Roskam and Anemaat73] are examples of the classic all-in-one MDO tools. At the other end of the spectrum are trade study environments such as Phoenix Integration’s ModelCenter [Reference Malone and Woyak74] or NASA’s OpenMDAO [Reference Gray, Moore and Naylor75].

The modern industrial MDO era is largely tied to existence of reliable commercial software. For an aircraft systems design perspective, the late 1990s reflected a paradigm shift from monolithic codes (in-house equivalents to Raymer’s RDS or NASA’s FLOPS) to those employing a flexible trade study environment (such as JMP and/or ModelCenter). Whereas many monolithic codes goals were to develop a classical wing-loading/thrust-loading plot, albeit using a higher-fidelity basis, flexible trade study tools enable engineers to make ‘run-time’ choices and explore high-dimensionality design spaces in a fundamentally different manner.

Since 2000, major airframe primes have showcased their in-house MDO capability as applied to different system design challenges. These published studies move far beyond classic wing-loading/thrust-loading constraint plots. In 2004, Carty & Davies at Lockheed Martin Skunkworks showcased the use of ModelCenter as a fusion environment linking numerical tools directly to CAD interface; see Fig. 13. Boeing showcased similar capability in 2006; Vandenbrande [Reference Vandenbrande, Grandine and Hogan57] emphasised the critical need for design-space-exploration as an integral part of the MDO process; see Fig. 14. High-speed vehicles have been an area of particular interest, as they are much more coupled designs than subsonic transports. In 2005, Raytheon Aircraft showcased MDO tools applied to low-boom supersonic business jet design [Reference Aronstein and Schueler77]. One year later, Raytheon Missile Systems demonstrated how they included aero-thermal effects in their ModelCenter-enabled MDO sizing tool used to screen high-speed weapons concepts [Reference Cunnington, Takahashi and Bays78].

Figure 13.

Aircraft design employing CAD integrated with analysis using ModelCenter [Reference Carty and Davies76].

Figure 14.

Design space exploration using statistical methods [Reference Carty45].

As basic MDO frameworks matured, we see government laboratory lead system-design MDO studies expanding existing models to include new disciplinary elements. For example, the EU Clean Sky joint-technical programme brought together government laboratory researchers, EU based industrial partners as well as academic researchers. In such an endeavor, MDO frameworks provide an essential collaborative function between the many developers and users. Clean Sky related publications include government/academic teams [Reference Hecken79] as well as small business [Reference Schneegans80]. In the United States, NASA sponsored several MDO-centric efforts including N + 2 supersonic airliner [Reference Magee81] and environmentally responsible aircraft studies [Reference Schutte and Mavris82]. Also in the United States, military-focused MDO programs have included the Air Force Research Laboratories led Efficient Supersonic Air Vehicle (ESAV) effort, which produced MDO papers led by major airframers [Reference Davies, Stelmack, Zink, De La Garza and Flick83], government staff [Reference Alanyak and Allison84] as well as university researchers [Reference Allison, Morris, Schetz, Kapania, Watson and Deaton85].

3.4 Collaborative design

The complexity of aircraft design has grown significantly over the past 120 years. While the Wright brothers’ aim was demonstrating controlled flight, the requirements for current commercial or military aircraft have risen due to a continued interplay between user demand and technology capability. Good flight performance is no longer the only objective, the aircraft must meet increasing noise and emission requirements, economic constraints, passenger appeal, maintainability, repairability, safety, etc. Therefore, solving current and future aircraft design problems require a collaborative effort from engineers, scientists, economist, aircraft operators, supply chain partners, government agencies, regulatory agencies, etc. These disciplines must be considered equally throughout the design process, as outlined in section 3.3, and all design decisions must be underpinned by top-level requirements (TLR). In multi-disciplinary design, disciplines are still standalone but interact with each other through data exchange. Some design decisions are still made within disciplines if this is considered reasonable or practical at the given stage of the design process, without necessarily linking them to TLR. In the future disciplines will become more integrated in a transdisciplinary environment where disciplinary boundaries become blurred, and all design decisions are driven by TLR.

System engineering has characteristics of art and science because good system engineering requires the creativity and knowledge of systems engineers, but it also requires systems management and the application of a systematic disciplined approach. The traditional design methodology is the sequential approach where specialists working ‘in series’ with feedback loops to previous steps. Lack of communication among the specialists cause incorrect assumptions to be adopted and the main system parameters are not monitored in real-time. This method reduces the opportunity to find interdisciplinary solutions and to create system awareness in the specialists.

Concurrent or Collaborative Engineering is as an alternative to the classical approach and provides better performance by taking full advantage of modern information technology (IT) (Fig. 15). Experts from various disciplines are co-located, either physically or virtually, and can communicate in real-time. As a result, many disciplines are involved in the design process and the concurrent approach has been proven to be particularly effective for complex systems design [Reference Fortescue, Swinerd and Stark86].

Figure 15.

Collaborative Design process.

A definition of concurrent engineering adopted by the European Space Agency (ESA) is [Reference Bandecchi, Melton and Ongaro87]:

Concurrent or Collaborative Design requires an IT environment and information system, a so-called collaborate design facility (cdf), allowing domain experts to communicate, share data, conduct design analysis, display results, etc. A CDF can be a centralised facility but will likely be connected to other sites in the organisation. Key elements of a CDF, include process, multidisciplinary team, integrated data model, appropriate facility and software infrastructure, will be discussed.

3.4.1 Background and history

Some attempts at concurrent engineering in aerospace and defense began in 1980s. In 1988, the Integrated Process Laboratory at the Concurrent Engineering Research Center (CERC) was established at West Virginia University by the Defense Advanced Research Projects Agency (DARPA) to promote CE to the U.S. industry. A survey showed that the major impetus in moving to a CE environment was seen to be the potential of reduction in overall costs and design time, as a pathway to be competitive and to improve product quality. It also showed that most pressing need was to foster a teamwork environment and that a greater leverage exists in teamwork and process improvement [Reference Lawson and Karandikar89].

The first CDF with full features, called the Project Design Center (PDC), was opened by the Jet Propulsion Laboratory (JPL) in 1994 [Reference Smith90]. The PDC provides a facility, with multiple rooms, for design teams to use to conduct concurrent engineering sessions. The Aerospace Corporation had developed the process and the tools for CE almost at the same time and they had been successfully applied to several programs. Based on this experience, the JPL contracted the Aerospace Corporation to develop CEM processes and tools for the PDC. The Concept Design Center (CDC) was established by the Aerospace Corporation in 1997 to enhance support to its customers by providing a process for bringing together the conceptual design capabilities and experts [Reference Aguilar, Dawdy and Law88].

In the European space industry, concurrent engineering was applied to spacecraft design from the beginning of 1990. The first establishment was the Satellite Design Office (SDO) at DASA/Astrium, with the cooperation of the System Engineering (SE) group at the Technical University of Munich. An experimental Concurrent Design Facility (CDF) was established at the ESA Research and Technology Centre (ESTEC) at the end of 1998 and used to performance the assessment of several missions. It is primarily used to assess the technical, programmatic, and financial feasibility of future space missions and new spacecraft concepts. Additionally, the ESA CDF is also used for many other multi-disciplinary applications, such as payload instrument preliminary design, System of System (SoS) architectures, space exploration scenarios, etc. [Reference Koning93]. The facility was upgraded in 2008 and has since delivered a design time reduction by a factor of 4 and a cost reduction by a factor of 2 [Reference Koning93]. Some Collaborative Design facilities in operation are shown in Fig. 16.

Figure 16.

Concurrent design facilities.

3.4.2 Key Elements in collaborative design

As shown in Fig. 17, a typical CDF includes the following key elements: a process, a multidisciplinary team, integrated design models, an IT facility, and a software infrastructure [Reference Bandecchi, Melton, Gardini and Ongaro91]. These elements are described in more detail, based on the CDF implementations at ESA/ESTEC. However, the concept of a CDF can be applied to solving a complex design problem as the only difference is in the design models which have to be specific.

Figure 17.

Elements of a concurrent design facility [Reference Braukhane92].

3.4.2.1 Process

Complex systems have many interdependencies between components. This implies that the definition and evolution of each component has an impact on other components and that any design change will propagate through the system. Early assessment of the system-level impact of changes is essential to ensure that the design process converges on an optimised solution.

The process defined way the team will work together. It typically starts with a preparation phase in which the team and the customer meet to refine and formalise the mission requirements, define the constraints, identify design drivers and estimate the resources needed to achieve the objectives. Then a kick-off takes place and the design process starts. It is conducted in a number of sessions in which all domain experts participate. This is an iterative process that addresses all aspects of the system design in a quick and concurrent fashion. One key factor is the ability to conduct a process that is not dependent on the path followed. At any stage it must be possible to take advantage of alternative paths or use ‘professional estimates’ to ensure that the process is not blocked by lack of data or lack of decisions.

3.4.2.2 Multidisciplinary team

Human resources are the most important and crucial element. A fundamental part of the Collaborative Design approach is to create a multi-disciplinary team representing all relevant domains that performs the design work in real-time. The collective approach, the collaborative environment, the intense and focused effort and clear objectives are all essential elements that contribute to personal motivation.

For each discipline representation is created within the facility and assigned to an expert in that technical domain. Each position is equipped with the necessary tools for design modeling, analysis and data exchange. The choice of disciplines involved depends on the level of detail required and on the specialisation of the available expertise. The number of disciplines may be limited initially, especially in the first experimental study, to avoid extended debate and to allow a fast turn-around of design iterations.

3.4.2.3 Integrated design models

The design process is model driven, using information derived from the collection and integration of the tools used by each specialist for their domain, that allows generic models of various mission/technological scenarios to be characterised for the study being performed (Fig. 18). A parametric approach supports fast modification and analysis of new scenarios, which is essential for the real-time process. It acts as a mean to establish and fix the ground rules of the design and to formalise the responsibility boundaries of each domain. Once a specific model is established it is used to refine the design and to introduce further levels of detail.

Figure 18.

Conceptual model of the mission and spacecraft design process [Reference Koning93].

Each model consists of an input, output, calculation, and results area. The input and output areas are used to exchange parameters with the rest of the system (i.e. other internal and external tools and models). The calculation area contains equations and specification data for different technologies to perform the actual modeling process. The results area contains a summary of the numeric results of the specific design to be used for presentation during the design process and as part of the report at the end of the study.

3.4.2.4 IT facility

The team of domain experts meet in the Concurrent Design Facility (CDF) to conduct design sessions. The accommodation generally comprises a design room, a meeting room and project-support office space (Fig. 19). The equipment location and the layout of the CDF are designed to facilitate the design process, the interaction, the collaboration, and the involvement of the domain experts. The facility is equipped with computer workstations each dedicated to a technical discipline. A central multimedia wall, supporting two or three large projector screens, can show the display of each workstation, so that the specialists can present and compare design options or proposals and highlight any implications imposed on, or by, other domains.

Figure 19.

The ESA/ESTEC concurrent design facility layout [Reference Koning93].

3.4.2.5 Software infrastructure

An infrastructure to implement the Collaborative or Concurrent Design Facility hardware outlined above requires software tools for the generation domain models with a means to exchange data between models in real time, a means to incorporate domain-specific tools for modeling and/or complex calculations, a documentation-support system, and storage capability (Fig. 20). The infrastructure must allow domain experts to work remotely from other sites, and exchange information easily between the normal office working environment and the CDF environment.

Figure 20.

Architecture of the software model [Reference Koning93].

Domain-specific tools by each expert have to be integrated into the infrastructure of the facility. This aspect of the infrastructure can be improvement over time and will enhance the concurrency of the design process. In some specific cases, e.g. the interface between domain specific tools, it is more convenient not to use the centralised data exchange, but rather to create a direct interface between applications. An example of such an interface is the transfer of geometrical 3D data of the design configuration to the simulation system, which uses geometric models for visualisation.