CIRP Annals - Manufacturing Technology 62 (2013) 99–102
Contents lists available at SciVerse ScienceDirect
CIRP Annals - Manufacturing Technology jou rnal homep age : ht t p: // ees .e lse vi er . com /ci r p/ def a ult . asp
Maintaining design intent for aircraft manufacture Mark A. Price a, Trevor T. Robinson a, Danielle Soban a, Adrian Murphy a, Cecil G. Armstrong a, Roisin McConnell a, Rajkumar Roy (1)b,* a b
School of Mechanical and Aerospace Engineering, Queens University Belfast, UK Manufacturing and Materials Department, Cranﬁeld University, Bedford, UK
A R T I C L E I N F O
A B S T R A C T
Keywords: Design Manufacturing Computer aided design (CAD)
Design and manufacture of aircraft requires deep multi-disciplinary understanding of system behaviour. The intention of the designer can get lost due to the many changes occurring to the product and the inability of the methods and tools used to capture it. Systems engineering and optimisation tools underpin industrial approaches to design, but are not without issue. The challenge is to ﬁnd a route from concept to manufacture which enables designers to maintain their original intent. The novelty in this work is that the parameterisation used to build the CAD model reﬂects the manufacturing capability, ensuring design intent is maintained from concept to manufacture. ß 2013 CIRP.
1. Introduction The introduction of new aircraft to market requires that ideas are transformed into operational products that meet market needs. This simple statement belies a complex, multi-faceted ‘‘design process’’ that typically includes sophisticated engineering analysis as well as advanced manufacturing methods. A prime function of this design process is getting the geometric deﬁnition of the product correct so that a product that meets market/performance requirements can be manufactured. Design has many deﬁnitions which emerge from different professions and industrial sectors. Many relate to design as being the act of creating the geometric description of thing to be made. Anderson and Raymer have two with respect to aircraft: ‘‘The intellectual engineering process of creating on paper a ﬂying machine that either meets certain requirements and performance objectives, or explores new concepts, technologies and innovations ’’ and ‘‘Creating the geometric description of a thing to be built ’’. Thus, design is fundamentally about deﬁning geometry, ultimately for manufacture, and Raymer’s deﬁnition is a key one as manufacture is rooted in it. But considering manufacture while designing a product is very difﬁcult. Today’s design processes are complex and even deﬁning the geometric representation of the product in such a way that it can evolve through the design process is arduous, with each specialism within the design process having different requirements from the geometric model. As a result, current industrial practice is to build multiple geometric models for a component just to represent the different aspects of shape. Each is used for a different application with little consideration given to the process used to create them, or their appropriateness as the design evolves, which ultimately depends
* Corresponding author. Tel.: +44 1234 750111; fax: +44 1234 754605. E-mail address: [email protected]
ﬁeld.ac.uk (R. Roy). 0007-8506/$ – see front matter ß 2013 CIRP. http://dx.doi.org/10.1016/j.cirp.2013.03.124
on how geometric shape is parameterised. Systems engineering  has become the main framework the aircraft industry follows as its design process. But this approach does not specify how geometric models of the design should be developed, or how the design moves from one phase to another. It also does not indicate how to control and manage all the variables that appear and disappear as a design evolves. Some design processes now deﬁne models based on parameters suitable for optimising shape, but very few select parameters based on the key requirement of embedding manufacturability in the representation. Several earlier studies considered frameworks and design systems using feature based approaches to support integrated manufacturing [4–6]. However these were typically focused on the deﬁnition and use of features themselves and how they can be used to support design decisions, mainly around machining, and process planning. These provided notable progressions in capability, but the complexity of assembly and the corresponding manufacturing processes have stretched these and control of design intent and requirements is again an issue for large complex systems. At the end of the design process a shape which can be manufactured must exist, but the question of how to create model geometry and manage parameters in a useful way is undeﬁned by modern processes. This work takes a novel approach by selecting the CAD modelling strategy and CAD parameters which reﬂect the capability of the manufacturing processes to be used. This allows the design to be optimised in terms of what it is possible to manufacture as opposed to a notional/optimal parameterisation where the design has to ultimately be approximated for cost effective manufacture. The objectives of this work are to demonstrate: the limitations of the technology which underpins the modern design process that the CAD parameterisation must reﬂect the manufacturing capability to embedding design intent in the design process
M.A. Price et al. / CIRP Annals - Manufacturing Technology 62 (2013) 99–102
2. Design intent Design intent is a term widely used to describe the way in which design geometry is modelled, but it appears to be often used without a clear understanding of its meaning. Deﬁnitions of design intent mainly stem from geometric modelling aspects, in particular regarding data exchange via STEP [7–9] and more recently focused on geometric design intent. Li et al.  has identiﬁed the importance of maintaining design intent in a CAD model, where the purpose is to control the properties of shape, and is expressed through relationships between the geometric entities. Other observations [9,11,12] on design intent note aspects as expressions of the process the designer uses to create a model. In particular Pratt and Anderson  note that the geometry alone does not carry intent, and that most importantly design intent is lost with the loss of the model building process. In most of the existing work the focus is capturing design intent within the STEP format, which speciﬁes how to translate the process used to create the shape (the order of features used). It is interesting to note that despite a clear deﬁnition of design intent provided originally in the STEP work it has been reinterpreted many times, or different abstractions given that are relevant to a particular project. Mun et al.  capture Anderson’s original deﬁnition succinctly as ‘‘the functional requirements provided by the customer: that is a set of geometric and functional rules which the ﬁnal product has to satisfy’’. The literature addressing the capture of design intent and model transfer has many looser variants. None of these examples or deﬁnitions of design intent completely coincide, and all use broad terms such as ‘‘intellectual arrangement’’ or ‘‘creative objectives’’. These concepts are useful when considering a holistic view and consideration of understanding or evaluating the original intent of an already completed design. But they are less useful to guide how a design should evolve. A further consequence of the failure to capture design intent robustly is that many models are created for the same component for different disciplinary analyses. In this case, the design intent is repeatedly reinterpreted with little consideration as to whether it is appropriate or not. In fact, as section 4 shows, different CAD parameterisation strategies produce different optimised structures and these can be highly dependent on the manufacturing process to be used. Here the functional variant of the deﬁnition of design intent is the basis for this work. A function is a process, activity or task that occurs over time and produces recognisable results as per IDEFO. The functional architecture of a system describes the product/ system in terms of what it does. But the architecture alone does not fully capture design intent as it does not specify the tools or methods used to deﬁne the geometry, nor does it provide a measureable approach model parameterisation. Hence design intent is deﬁned for a general system as ‘‘The set of functions which the system was designed to deliver in the anticipated operating environment’’. Having a clear deﬁnition of design intent it is easier to move on and consider the issues of model parameterisation and the model building process. 3. Modelling approaches In modern CAD systems the parameterisation is introduced by adding CAD modelling features, e.g. an extruded boss. When creating a CAD model the designer chooses the CAD modelling features to use to deﬁne the shape. It is the process used to create the model that embodies the design intent . 3.1. CAD modelling procedures and features The number of feature creation tools available in a CAD system makes the task of identifying the best modelling strategy extremely difﬁcult. A simple example is a square, which can be created in a CAD system using several different methods, Fig. 1. It
Fig. 1. Deﬁning a square using different parameterisations.
can be deﬁned based on two corner points Fig. 1a, a centre and a diagonal Fig. 1b, a baseline can be swept Fig. 1c, deﬁning the four corner points and using them to deﬁne the lines which bound the square Fig. 1d. Each one of these may tell the intent of the designer, or they may simply reﬂect the creativity of the draftsman or even just the data most accessible at the time. The real evolution of the design to that point may be simply a rectangle which for the current design iteration has equal dimensions for breadth and depth, Fig. 2e. As well as adding parameters, features also impose constraints on shape. A practical example is an extruded boss which many CAD systems constrain to have a constant cross section along its length. To embed manufacture in design, design intent should help clarify which model building approach best reﬂects the manufacturing capability.
Fig. 2. Deﬁning a cylindrical tube.
3.2. Parameters Anderson expresses that ‘‘the designer’s choice of parameterisation and constraint schemes constitutes an important part of what is known as design intent’’ . Without having a clear grasp of how the design is deﬁned subsequent decisions downstream are more difﬁcult to make and their consequences more uncertain, creating a disjoint between design intent and the resulting design. Fig. 2 shows a cylindrical tube representing a fuselage. In Fig. 2a the inner and outer radius of the tube are both deﬁned as model parameters. Although they deﬁne the shape of the proﬁle this is not the optimum choice of parameters if wall thickness is to be controlled. Fig. 2b represents wall thickness as a parameter which can be controlled directly, but the parameter on inner radius means that varying wall thickness, without a corresponding change of inner radius, will change the outer radius, which for this design may be required to stay constant. Fig. 2c allows the wall thickness to be varied directly in an optimisation, and will keep the outer radius constant, while varying wall thickness with variation happening on the inner radius. Thus the choice of parameterisation effects how the model shape is controlled. This example demonstrates the effect a badly deﬁned parameterisation can have on a design. In reality there may be different teams of engineers working to deﬁne each dimension, in which case a transformation is needed between each of the parameterisations described, while ensuring each is consistent and that incompatible solutions do not result from each. The wrong choice of parameterisation may result in the inability to capture certain functions and errors needed to make informed design decisions. With this level of complexity attached to the parameterisation of a cylinder it is easy to extrapolate and see how complex the problem is for an assembly of thousands or millions of parts. Once the focus is shifted outside geometric description, other design information needs to be captured. For example, with regard to manufacture of the cylindrical tube, it can be made by extruding a hollow circular section, or from several curved extrusions making up the circular cross-section, or several short cylinders, or a single ﬂat sheet rolled and welded and so on. Which manufacturing process is to be used should clearly have an impact on how the CAD
M.A. Price et al. / CIRP Annals - Manufacturing Technology 62 (2013) 99–102
model is built and which parameters are appropriate to control and constrain the shape. For example where multiple sections are being joined together it is essential to ensure that the inner and outer radius values for adjacent sections are compatible. As design is an iterative process the shape, materials etc. will gradually change as the design progresses, so what started as a simple cylinder may no longer be identiﬁed as such in the ﬁnal detailed design when windows, doors etc. are added (Fig. 3). The manufacturing process capability may mean that wall thickness does not need to be constant, in which case a design freedom is available which the parameterisation in Fig. 2 is not capable of representing.
Fig. 3. Detailed representation of a fuselage.
manufacturing process capability so that any optimisation/ analysis feeds the design model directly. To account for this in the model building process the concepts of dimensional addition and detail insertion are used for the creation of geometric models . Dimensional addition involves deﬁning a model at one level of detail, and building the ﬁnal component model through a series of operations which add detail and dimension. Examples include extruding a proﬁle in a direction or along a curve, or sheet thickening operations where sheet bodies are converted to solid bodies by sweeping material outwards. Detail insertion involves adding features of localised geometric detail as features deﬁned relative to higher level geometry features. In other words, a beam is represented by its neutral axis and cross-sectional shape, and a thin plate is represented by its shape and a thickness. A hole feature might then be placed relative to the neutral axis at a point along the beam. The beam neutral axis is placed relative to its higher level entities in the assembly. In the case of a stiffener in a fuselage it would be relative to the main fuselage neutral axis. Interfaces in assemblies of course are also represented (Fig. 5).
4. How to generate models Clearly the parameterisations described for Figs. 1 and 2 do not capture the information needed to take design through to manufacture, because the parameterisations do not account for the capability of the manufacturing process. The intention of the designer is lost or obfuscated both in the natural evolution of the process, and the way the design tools have been used to implement this. Consequently, to ensure intent is embedded within the CAD model an appropriate parameterisation is needed which reﬂects the manufacturing process, and a model construction method which directly relates to the structural function of the component is also required. Consider ﬁrst the issue of parameterisation. Fig. 4 shows two methods of parameterising a stiffened panel, reﬂecting different manufacturing process. The top parameterisation could represent the case where the three regions in dashed lines are to be machined out of a block of material. The bottom parameterisation could represent where two stiffeners are attached to a sheet of material. In their current state both parameterisations represent the same shape, however selecting the bottom parameterisation constrains the base sheet thickness to be constant, whereas machining out the material allows D1, D2 and D3 to assume different values. Even this imposes a constraint that the top surface of the thin sheet is ﬂat, whereas this need not be the case for all processes. Either can deﬁne the shape, but the one chosen should reﬂect the
Fig. 4. Parameterisation of a stiffened panel.
Fig. 5. Idealisation/abstract representation of a stiffener on a plate.
Applying this to the panel in Fig. 4, if the panel is created by dimensional addition, where a sketch of the proﬁle is swept in a straight line to create the solid component as per the bottommost sketch. Sweeping the sketch in this way also enforces a constant section along the component. So how does this really help? Consider a simple stiffening element consisting of a plate with a Z shaped stiffener. Fig. 6 shows a number of different methods for ﬁxing the stiffener to the thin sheet. Attempting to use the same parameterisation for each type would be inappropriate. The optimum design locally to the fastener (Fig. 6a) will be different from that for say the laser welded specimen (Fig. 6d), and no single parameterisation will provide the appropriate level of shape control to reﬂect all of the different options. In fact, detailed studies on optimising these structures [14–18] have shown that the analysis methods typically used in design could not distinguish sufﬁciently between the different manufacturing processes. Yet detailed simulations backed up with testing demonstrated that signiﬁcant weight savings of the order of 15% could be achieved used either of the welding processes [15,16].
Fig. 6. Joining a stiffener to a panel.
M.A. Price et al. / CIRP Annals - Manufacturing Technology 62 (2013) 99–102
Clearly, if a different parameter set is used to deﬁne the solid model then switching between the variants is difﬁcult. In this work a regenerative approach is used where when the choice of manufacturing process changes the model is regenerated with the appropriate parameter set to deﬁne it, but yet is consistent with the design decisions which have already been made. This is a more controlled rollback than via the typical feature tree in CAD. The ability to control the way in which parameterisations and processes are deﬁned is key in this work, and facilitated through the hierarchical abstract model, and sweep/thickening operations. It is then easy to switch between any one of these representations. It is interesting to note that a better design was found in the laser specimens moving the stiffener to the middle of the pad. This new approach treats this as per Fig. 6d except the neutral axis of the stiffener is shifted slightly in position. It is important that a situation does not occur where the parameterisation selected for the model dictates design, Fig. 7 (e.g. the manufacturing process which has to be used). Using a parameterisation which cannot represent the level of shape control available in the manufacturing process will result in conservative designs which contain material not necessary for the function of the product and which could have been manufactured out. For an entire aircraft the cost of this extra weight due to a badly selected parameterisation could be substantial, and could have a signiﬁcant impact on the success of the product in the marketplace.
Fig. 7. Switching between different manufacturing paradigms (composite sandwhich v metallic semi-monocoque) requires an agile approach to parameterisation and model generation.
A major beneﬁt to this function based approach to modelling is that design intent becomes measureable  provides a measure for the effectiveness of the parameterisation of a CAD model for optimisation purposes. Design intent could be measured by comparing the optimum shape the CAD model can assume based on its parameterisation to the shape which can be manufactured. This simple measure would indicate how well the manufacturing process is represented in the design model. 5. Discussion Enforcing a design intent that tailors the parameterisation of the geometric model to the capability of the manufacturing process is the only method of ensuring the ﬁnal design is the optimum ﬁnal product. Yet in a product as complicated as an aircraft this is not easy to do. Systems engineering provides a requirement analysis process and requirements ﬂow down from the system top level to each subsystem as needed, as it does for the functional ﬂow down. But requirements and functional allocations do not prescribe which parameters to use as these often evolve with the design. It has been shown here that selecting a parameterisation to reﬂect manufacturing capability is key to maintaining design intent. Where the parameterisation is too ﬂexible (too many parameters) the resulting design will not be easily manufactured using
the selected process. The consequence is a ﬁnal component which deviates from the design, or a new manufacturing process being adopted which could add signiﬁcant cost. Where the parameterisation is too inﬂexible (too few parameters) the resulting design does not take into account the versatility of the selected manufacturing process, and a design is created which is not as optimised as could be, and which is ultimately heavier than necessary, increasing operating cost. Hence, there is a real need to select modelling strategies and parameterisations which are capable of reﬂecting the function of the component being designed and the manufacturing process being used to produce the component.
6. Conclusions This work shows that to maintain design intent throughout the design process requires a systematic approach to both model building and model parameterisation. By using CAD modelling procedures and parameterisations that reﬂect the manufacturing process capability, design intent is more naturally embedded in the design models.
References  Anderson JD (1999) Aircraft Performance and Design, McGraw-Hill, UK.  Raymer DP (1999) Aircraft Design: A Conceptual Approach, 3rd ed. AIAA Education Series Textbooks, Reston, VA, USA.  NASA. (1995) NASA Systems Engineering, Handbook, NASA, Hanover, MD, USA.  Kimura F, Suzuki H (1989) A CAD system for efﬁcient product design based on design intent. CIRP Annals – Manufacturing Technology 38/1:149–152.  Dong X, DeVries WR, Wozny MJ (1991) Feature-based reasoning in ﬁxture design. CIRP Annals – Manufacturing Technology 40/1:111–114.  Harutuniana V, Nordlundb M, Tatea D, Suha NP (1996) Decision making and software tools for product development based on axiomatic design theory. CIRP Annals – Manufacturing Technology 45/1:135–139.  Pratt MJ, Anderson BD (2001) A shape modeling applications programming interface for the STEP standard. Computer-Aided Design 33:531–543.  Mun D, Han S-H, Kima J, Ohb Y (2003) A set of standard modeling commands for the history-based parametric approach. Computer-Aided Design 35:1171– 1179.  Pratt MJ, Anderson BD, Ranger T (2005) Towards the standardized exchange of parameterized feature-based CAD models. Computer-Aided Design 37:1251– 1265.  Li M, Lanbein FC, Martin RR (2010) Detecting design intent in approximate CAD models using symmetry. Computer Aided Design 42/3:183–201.  Iyer GR, Mills JJ, Barber S, Devarajan V, Maitra S (2006) Using a context-based inference approach to capture design intent from legacy CAD. Computer-Aided Design & Applications 3/1-4:269–278.  Chen X, Gao S, Yang Y, Zhang S (2012) Multi-level assembly model for topdown design of mechanical products. Computer-Aided Design 44:1033–1048.  Price M, Raghunathan S, Curran (2006) An integrated systems engineering approach to aircraft design. Progress in Aerospace Sciences 42/4:331–376.  Murphy A, Price M, Lynch C, Gibson A (2005) The computational post buckling analysis of fuselage stiffened panels loaded in shear. International Journal of Thin-Walled Structures 43:1455–1474.  Murphy A, Lynch C, Price M, Gibson A (2006) Modiﬁed stiffened panel analysis methods for laser beam and friction stir welded aircraft panels. IMechE Journal of Aerospace Engineering 220:267–278.  Murphy A, Price M, Curran R, Wang P (2006) The integration of strength and process modelling of friction-stir-welded fuselage panels. AIAA Journal of Aerospace Information & Communications Technology 3/4:159–176.  Murphy A, McCune W, Quinn D, Price M (2007) The characterization of friction stir welding process effects on stiffened panel buckling performance. International Journal of Thin Wall Structures 45/3:339–351.  Curran R, Price M, Raghunathan S, Benard E, Crosby S, Castagne S, Mawhinney P (2005) Integrating aircraft cost modelling into conceptual design. Concurrent Engineering Research and Applications (CERA) 13(4):321–330.  Robinson TT, Armstrong CG, Chua HS (2013) Determining the parametric effectiveness of CAD models. Engineering with Computers 29/1:111–126.