How do I generate or create feasible designs?

7.1 GENERATING THE “DESIGN SPACE,” A SPACE OF ENGINEERING DESIGNS

How do we generate or create actual designs? We start by building a design space, an imaginary intellectual region of design alternatives that contains all of the potential solutions to our design problem. A design space is a useful notion that conveys a feel for the problem at hand: A large design space suggests a design domain with a large number of acceptable designs, or a design problem with a large number of design variables. While we can often look at a domain and intuit something about its design space (e.g., auto and building designs occupy very large design spaces), it is not clear how we identify a design space for unfamiliar or new devices. We now introduce the morphological chart as a formal tool for generating design spaces and for generating within those spaces a population of designs that perform the functions we specify. After that we will look at analogical thinking, another approach for generating design alternatives, and then offer one verbal and two graphical tools for coming up with designs in team-based activities.

7.1.1 Defining a Design Space by Generating a Morphological Chart

A morphological chart (aka a morph chart) is a matrix in which the leftmost column is a list of all of the principal functions that our design must perform and also some of the key features it must have. The list should be of a manageable size, and all of the entries should be at the same level of detail to help ensure consistency. Then, across from each of the functions or features, we list each of the different means of realizing the function or feature that we can think of. We strongly encourage separating functions from key features, for several reasons. First, we know that our design must be fully functional to satisfy our client's requirements. By putting all the functions together on the morph chart, we know we have addressed them all. The second reason we encourage separating functions from features is that a morph chart can quickly become rather large, and we may lose track of or confuse functions with key features. If we separate function from features at the outset, we can easily create two “design space” models in two separate charts, if necessary.

If we listed all the functions for the beverage container problem and arrayed the means corresponding to each to the right of each entry, we'd get the morph chart shown in Figure 7.1. We see that some functions have more means than others, for example, the function Contain Beverage has four means, while Resist Forces has only two. When we see a very small number of means this suggests that either we have a small design space (i.e., limited choices) or we have not fully explored the available design space.

We start building conceptual designs from the morph chart by noting that any feasible design must be functionally complete: every function, listed in the leftmost column must be achieved by that design. So we assemble designs by choosing one means from each row, and combine them into a functional design concept or scheme. Thus, we see in Figure 7.2a that one feasible design for the new juice container is a heat-sealed bag with a tear corner, thick walls, and a distinctive label, and another is a bottle with a twist top, made of a flexible material and with a distinctive shape.

Figure 7.1 A morphological (“morph”) chart for the juice container design problem with functions listed in the leftmost column. The means by which each can be implemented are arrayed along a row to each entry's right.

Figure 7.2 The morphological (“morph”) chart for the juice container design problem (Figure 7.1) is used to show (a) two feasible design alternatives whose means are dark and light shaded, and (b) two infeasible combinations whose means are also dark and light shaded.

The design generation method we have just described makes the morph chart into a spreadsheet with which we can “calculate” the number of potential designs. How many potential designs are there in that morph chart, that is, just how big is our design space? The answer reflects the combinatorics that result from combining a single means in a given row with each of the remaining means in all of the other rows. Thus, for the beverage container morph chart of Figure 7.1, the number of design alternatives could be as large as 4 × 5 × 6 × 2 × 3 = 720.

While it seems that the design space for this simple example has suddenly become very large, it is important to recognize that not all of the combinations allowed by our combinatorial arithmetic are valid combinations, that is, not all of these 720 combinations are feasible designs. For example, we can see (Figure 7.2b) that we can't really design a bag with a zipper or a can with an unfolding corner!

Thus, our morphological chart provides both a tool to develop a design space and create design alternatives, and it provides an approach to prune that design space by identifying and excluding infeasible, incompatible alternatives. We exclude infeasible alternatives by, again, applying interface constraints, as well as physical principles and plain common sense.

We can use the morph chart to include key features as well as functions. In the juice container, for example, we might include a set of entries related to the materials we want to use, in which case we could distinguish between glass, plastic, Mylar, and cardboard. These features can help us understand our design space and conceptualize alternative designs, as well as generating more infeasible designs such as a glass bag.

There is a lot we can learn from our morph chart. Consider our problem with not having many means for Resisting Forces. The heart of this problem is that we need to consider resisting forces in more detail, distinguishing between Resisting Temperature and Resisting Shocks. At the same time, this may take us more deeply into particular designs than is appropriate at the conceptual stage. Once we have selected a concept, such as a bottle, we can increase its resistance to forces by thickening the walls, adopting appropriate structural elements, or wrapping it in a protective plastic. This shows us that it is important that we list functions (and features) at the same level of detail when we build a morph chart. Otherwise, we will find ourselves developing highly detailed designs at the conceptual stage, or still creating concepts even after we have settled on a scheme. Similarly, when doing a complex design task (e.g., designing a building), we don't want to worry about means for identifying exits or for opening doors while developing different concepts for moving between floors (e.g., elevators, escalators, and stairways).

We can also use morph charts to expand the design space for large, complex systems by listing principal subsystems in a starting column and then identifying various means of implementing each of those subsystems. For example, if we were designing a vehicle, we would need a subsystem Provide Power, which would have corresponding means like Gasoline, Diesel, Battery, Steam, and LNG. Each of these power sources is itself a subsystem that needs further detailed design, but we see we can use the morph chart idea to develop an array of subsystems to expand our range of design choices for a complex design. We might even choose to create a morph chart for some of these subsystems to help us appreciate the design choices implicit in our concepts and schemes.

7.1.2 Thinking Metaphorically and Strategically

As we consider our morph chart, it is natural to ask where the ideas for the means and implementations come from. In this and the following sections, we look at a number of “ways of thinking” along with tools and techniques, all intended to help us come up with lots of creative and appropriate ideas. The first of these builds upon ways of thinking and speaking that we use in our everyday life, and shows us how they can be applied to engineering design.

A metaphor is a figure of speech that is used to give depth or color to the description of an object or process by likening it to another, usually more familiar, object or process. For example, when we describe engineering education as drinking from a fire hose, we mean to suggest that engineering students are exposed to a lot of knowledge quickly and under great pressure. We use metaphors to point out analogies between two different situations, that is, to suggest that there are parallels or similarities in the two sets of circumstances. Analogies can be very powerful tools in engineering design as illustrated by perhaps the most often cited: the Velcro fastener was designed by someone who drew a direct analogy between plant burrs that seem to stick to everything on which they're blown and the connecting fibers of the fastener.

We might use symbolic analogies, as when we “plant” ideas or talk about objectives “trees,” because we are clearly drawing connections through some underlying symbolism. We can also stray into the realm of fantasy analogies by imagining something that is literally fantastic or beyond belief.

Fantasy analogies suggest another approach, sometimes called “thinking outside of the box.” We are not very far past the time when many of the technologies we take for granted were thought to be outrageous ideas that were beyond belief. When Jules Verne published his classic 20,000 Leagues Under the Sea in 1871, the idea of ships that could “sail” deep in the ocean was viewed as preposterous. Now, of course, submarines and seeing unfamiliar yet exciting life forms underwater are part of everyday experience. We cannot escape the idea that design teams might imagine the most outrageous solutions to a design problem and then seek ways to make such solutions possible. For example, airplanes that are invisible to radar were once considered far-fetched. The arterial stents used in angioplastic surgery (Figure 7.3) are also devices once thought to be impossible. Who would have believed that an engineering structure could be erected within the narrow confines of a human artery? And nowadays, of course, with the advent of nanotechnology, people are designing machines and devices just for use in human blood vessels.

Figure 7.3 This is a PALMAZ-SCHATZ™ balloon expandable coronary stent, that is, a device used to maintain arterial shape and size so as to allow uninhibited and natural blood flow. Note how this structure resembles the kind of scaffolding often seen around building renovation and construction projects. Photo courtesy of Cordis, a Johnson & Johnson Company.

The stent suggests still another aspect of analogical thinking, namely, looking for similar solutions. The stent is similar in both intent and function to the scaffolding erected to support walls in mines and tunnels as they are being built. Thus, the stent and the scaffold are like ideas.

We could invert this idea by looking for contrasting solutions in which the conditions are so different, so contrasting, that a transfer of solutions would seem totally implausible. Here we would be looking for opposite ideas. Fairly obvious contrasts would be between strong and weak, light and dark, hot and cold, high and low, and so on. One example of using an opposite idea occurs in guitar design. Most guitars have their tuning pegs arrayed at the end of the neck. In order to make a portable guitar, one clever designer chose to put the tuning pegs at the other end of the strings, at the bottom of the body, in order to save space and thus increase the guitar's portability.

In addition to finding similar and contrasting solutions, we recognize a third category. Contiguous solutions are developed by thinking of adjoining (or adjacent) ideas in which we take advantage of natural connections between ideas, concepts, and artifacts. For example, chairs prompt us to think of tables, tires prompt us to think of cars, and so on. Contiguous solutions are distinguished from similar solutions by their adjacency, that is, bolts are adjacent to nuts and are contiguous solutions, while bolts and rivets serve identical fastening functions and are thus similar solutions.

The kinds of metaphorical thinking we have just described are also related to a characterization of design in terms of two different kinds of thinking:

• We do divergent thinking when we try to remove limits or barriers, hoping to increase our store of design ideas and choices. Speaking metaphorically, we “think outside of the box” or “stretch the envelope” when we want to expand our space of design alternatives.
• We do convergent thinking when we try to narrow our design space to focus on the best alternative(s). Again speaking metaphorically, we want to “stay within our game” and “know where the boundary/goal is” so we can converge on a solution within known boundaries or limits.

Perhaps we can best sum up this dilemma of choosing between thinking styles with an adapted metaphor: Think outside the box, but stay within the physics!

7.1.3 The 6–3–5 Method

Because so much engineering design is done in team settings, it is useful to consider tools and techniques that are well suited to teams. We consider three of these team-based activities. As before, our intent is to generate a rich set of ideas that can help us explore the entire design space.

The first of these team-based design generation activities is known as the 6–3–5 method. Its name came from having six team members seated around a table to participate in this idea generation “game,” each of whom writes down three design ideas, briefly expressed in key words and phrases. The six individual lists are then circulated past each of the remaining team members in a sequence of five rotations of written (only) comment and annotation: verbal communication or cross talk is not allowed. Thus, each list makes a complete circuit around the table, and each member of the team is stimulated in turn by the increasingly annotated lists of the other team members. When all of the participants have commented on each of the lists, the team lists, discusses, evaluates, and records all of the design ideas that have resulted from a group enhancement of the individual team members' ideas in a common visualization medium (e.g., a blackboard or a projection screen).

We can generalize this method to the “m − 3 − (m − 1)” method by starting with m team members and using m − 1 rotations to complete a cycle. However, the logistics of ever lengthening lists written on increasingly crowded sheets of paper, and of providing tables that seat more than six, suggest that six may be a “natural” upper limit for this activity. (In an academic setting, we would prefer fewer than six—ideally no more than four—on a project team.)

7.1.4 The C-Sketch Method

The C-sketch method starts with a team seated around a table, with each member sketching one design idea on a piece of paper, and then proceeds as does the 6–3–5 method. Each sketch is circulated through the team in the same fashion as the lists of ideas in the 6–3–5 method, with all of the annotations or proposed design modifications being written or sketched on the initial concept sketches. Again, the only permissible communication is by pencil on paper, with discussion following only after a complete cycle of sketching and modifying (as in the 6–3–5 method) has been completed. Research suggests that the C-sketch method can become unwieldy with even five team members due to the crowding of annotations and modifications on a given sketch. However, the C-sketch method is very appealing in an area such as mechanical design because there is strongly suggestive evidence that sketching is a natural form of thinking in mechanical device design. Research has also shown that drawings and diagrams facilitate the grouping of relevant information (usually added in marginal notes), and they help people to better visualize the objects being discussed.

7.1.5 The Gallery Method

The gallery method is a third approach to getting team reactions to design idea sketches, although the communication cycles are done differently. In the gallery method, team members first develop their individual, initial ideas within some allotted time, after which all of the resulting sketches are posted on a corkboard or a conference room whiteboard. This set of sketches serves as the backdrop for an open, group discussion of all of the posted ideas. Questions are asked, critiques are offered, and suggestions are made. Then each participant returns to her or his drawing and suitably modifies or revises it, again within a specified period of time, with the goal of producing a second-generation idea. The gallery method is thus both iterative and progressive, and there is no way to predict just how many cycles of individual idea generation and group discussion should be held. Our only recourse would be to invoke the idea of a utility plot and apply the law of diminishing returns: We proceed until a consensus emerges within the group that one more cycle will not gain much (or any) new information, then we quit because have reached a saturation plateau.

Note that the C-sketch and gallery methods provide contexts for committing design thoughts to paper by sketching. In fact, design teams do a variety of sketching and drawing activities, for a variety of purposes, and using a variety of technologies. We will discuss this (and show examples) in Chapter 9 and Appendix II, but we emphasize that the C-sketch and gallery methods need only rudimentary sketches, so a designer need not be an artist to be a visual thinker.

7.1.6 Guiding Thoughts on Design Generation

It is worth remembering that design generation is an exciting, creative activity, but it is goal-directed creative activity: It is designed to serve a known purpose, not to search for one. The goal may be imposed externally, as is often the case in engineering design firms, or internally, as in the development of a new product in a garage. But, there is a goal toward which that creative activity is aimed.

It is also worth remembering that creative activity requires work. As Thomas Edison famously said, “Invention is 99% perspiration and 1% inspiration.” In other words, we have to be willing to do some serious work if we expect to be successful at generating design alternatives. Therefore, in order to do good, goal-directed design generation, we ask: Beyond the morph chart, beyond the team-based tools, what else can we do to generate design ideas? Or, how can we usefully navigate, expand or, if needed, contract our design space?

7.2 NAVIGATING, EXPANDING, AND CONTRACTING DESIGN SPACES

We began our discussion of design generation by proposing the morph chart as a formal tool for identifying spaces of designs that are populated by individual design alternatives. As we gain design experience, we find it more natural to think about design spaces and classes (not just individual designs) because we see commonalities across types of products or devices. Further, our experiences will increasingly turn toward designing more complex engineering systems, meaning that we'll have to design more subsystems and components, and we'll have to combine and connect these subsidiary individual designs. Thus, we offer a few thoughts about thinking effectively at the design space level.

7.2.1 Navigating Design Spaces

Large design spaces are complex because of the combinatorial possibilities that emerge when hundreds or thousands of design variables must be assigned. Design spaces are also complex because of interactions between subsystems and components, even when the number of choices is not overwhelming. In fact, one aspect of design complexity is that collaboration with many specialists is often critical because it is rare that a single engineer knows enough to make all of the design choices and analyses.

Two designed objects that have large design spaces are passenger aircraft (e.g., the Boeing 747), and major office buildings (e.g., Chicago's Sears Tower). A 747 has six million different parts, and we can only imagine how many parts there are in a 100-story building, from window frames and structural rivets to water faucets and elevator buttons. With so many parts, there are still more design variables and design choices. Yet, while both the 747 and the Sears Tower have very large design spaces, these devices differ from one another because their performances present different challenges and different constraints. Architect and structural designers of a skyscraper have far more choices for the shape, footprint, and structural configuration of a high-rise than do aeronautical engineers, who must design fuselages and wings within strict aerodynamic constraints. While a building's weight is important as its number of floors and occupants rise, and while high-rise buildings' shapes are analyzed and tested for their response to wind, they are subject to fewer constraints than are the payload and aerodynamic shape of aircraft.

A small or bounded design space, on the other hand, conveys the image of a design problem in which the number of potential designs is limited or small, or the number of design variables is small and they, in turn, take on values within limited ranges. Thus, the design of individual components of large systems often occurs within small design spaces. For example, the design of windows in both aircraft and buildings is so constrained by opening sizes and materials that their design spaces are relatively small. Similarly, the range of framing patterns for low-rise industrial warehouse buildings is limited, as are the kinds of structural members and connections used to make up those structural frames.

One of the complications of large design spaces stems from the fact that many design variables are highly dependent either on choices already made or on those yet to be made. We attack such complex design spaces by applying the idea of decomposition, or divide and conquer: divide (or break into pieces) a complex problem into subproblems that are more readily solved. Designs of airplanes, for example, can be decomposed into subproblems: the wings; the fuselage; the avionics; the tail; the galley; the passenger compartment; and so on. In other words, the overall design problem and space are broken into manageable pieces that are taken on one at a time. Indeed, the morphological chart is particularly suited to (1) decomposing the overall functionality of a design into its constituent subfunctions; (2) identifying means for achieving each of those subfunctions; and (3) supporting the synthesis, or composition and recomposition, of feasible design solutions.

7.2.2 Expanding a Design Space When It Is Too Small

We sometimes feel that our design space is too small, that we may not have enough options. There are things we can do to expand our design space, and they fall into various kinds of information gathering, including:

• Conduct literature reviews to determine the state of the art and identify prior work in the field. This includes locating and studying previous solutions, product advertising, vendor literature, as well as handbooks, compendia of material properties, design and legal codes, and so on. For example, The Thomas Register lists more than one million manufacturers of the kinds of systems and components used in mechanical design. Further, much more material is becoming available on the World Wide Web, although it is risky to assume that all information is both web-available and technically correct.
• Conduct a patent search to identify available technologies, to avoid “reinventing the wheel” and to leverage our thinking by building on what we already know about a still-emerging design. Patents are a kind of intellectual property: Patent holders identified by the U.S. Patent Office (USPTO) are credited for having invented new devices or discovered new ways to do things. The USPTO awards two kinds of patents: design patents on the form or appearance or “look and feel” of an idea; and utility patents for functions, that is, on how to do something or make something happen.
• Benchmark existing products to evaluate how well they perform.
• Reverse engineer devices to see how functions are performed and to identify alternate means of performing similar.

7.2.3 Contracting a Design Space When It Is Too Large

We often feel that our design space is too large, that we have too many options, so we need to prune or contract our space to make it more manageable. There are several pragmatic guideposts for narrowing a search space, including:

• Check for external constraints that affect the design. For example, make sure the design team's competence maps onto the design problem posed (e.g., it may be more comfortable designing tricycles than high-tech performance bikes). For another example, be sure that the manufacturing capabilities are available (e.g., a team should avoid designing a bike made of composite materials if the only available manufacturing facility forms and connects metals).

Invoke and apply constraints, in much the same way we did just above while assessing the presence of external constraints.

Freeze the number of features and behaviors being considered to avoid details that are unlikely to seriously affect the design at this point (e.g., the color of a bike or car is not worth noting in a design's early stages.)

Impose some order on the list, perhaps by harking back to data gathered during problem definition that suggests that particular functions or features are more important.

“Get real!” or, in other words, apply common sense to rule out infeasible ideas.

7.3 GENERATING DESIGNS FOR THE DANBURY ARM SUPPORT

We now return to following the two design teams working on the arm support for the CP-afflicted student at the Danbury School. Starting with the functions they had already identified, each team built a morph chart: Team A's, based on Table 6.2, is shown in Figure 7.4; and Team B's, based on Table 6.3, is displayed in Figure 7.5. In addition, although to different degrees, the teams did research on the availability of devices that were intended to serve the same functions for similar users. (It is worth noting, too, that teams in HMC's E4 design experience are routinely told that it is perfectly acceptable to recommend that the client buy an existing product provided they can identify an existing product that meets the client's objectives, performs the specified functions, and satisfies the problem's constraints. The teams are told this in part to let them know that this is a legitimate outcome, and in part to encourage the teams to do their research so as to avoid “reinventing the wheel”!) A comparison of the two morph charts reinforces our earlier observations about focused thinking because Team B's is a more bounded morph chart (and design space) reflecting a sharper, more concise list of functions. Similarly, the three passive functions (“Enable adjustability of …”) shown in Team A's chart, and their associated means, might once again be viewed as three simple articulations of subfunctions and means of the top-level function, “Enable adjustability.” In particular, if the overall objective is “Provide adjustability,” is this the right time in the design process to consider details of different kinds of adjustments?

Figure 7.4 Excerpts from Team A's morph chart for the Danbury CP arm support. Adapted from Attarian et al. (2007).

Figure 7.5 Team B's morph chart for the Danbury CP arm support. Adapted from Best et al. (2007).

We also note that both of these morph charts are very large, that is, they have very large numbers of possible combinations: 13,310 (i.e., 11 × 11 × 11 × 10) alternatives for the (partial) chart of Figure 7.4 and 7200 for the smaller morph chart of Figure 7.5. These are overwhelmingly large numbers of outcomes for this design problem, which suggests the teams should think strategically about grouping and reorganizing the functions and the resultant design alternatives.

Both teams followed similar approaches: they blended possibilities stemming from their morph charts with information gained from their research and with their own experience-based judgments and gut-level feelings. Team A developed three designs based on their morph chart, as can be seen in the marked-up version of their morph chart in Figure 7.6. They called those three designs: “Sling”; “Sliding Bars”; and “Support Arm.” We can also see from Figure 7.4 that there was a fair amount of overlap among their three designs, with several different means being used in more than one design. That may not be terribly surprising, given the nature of this particular design problem.

Team B took a less structured approach in which they decomposed the overall design into: a “structure component” that connected Jessica's arm to a structure that would maximize her range of motion and provide comfort and adjustability; and a “dampening component” that would minimize her involuntary contractions without restricting her voluntary motion or range of motion. The three designs were identified by Team B as “Dually-Hinged Structure with Rail,” “Dually Hinged Structure,” and “Ball and Socket Structure.”

Figure 7.6 A marked-up copy of the excerpts of Team A's morph chart for the Danbury CP arm support that shows how their three designs were assembled from their morph chart (Figure 7.4). Note that Team A often used two means to achieve a given function. The three designs are: (1) “Sling”; (2) “Sliding Bars”; and (3) “Support Arm.” Adapted from Attarian et al. (2007).

Some of the design sketches and drawings for the design concepts produced by the teams are shown in Figures 11.1 and 11.3 as part of our discussion of sketching, drawing, and prototyping. But is interesting to look ahead and see some of the most immediately tangible fruits (and joys) of successful conceptual design. Teams A and B selected the particular designs shown in Figures 11.1 and 11.3, respectively, after they applied their particular metrics to their corresponding objectives (see Figure 4.3 and Table 4.10).

7.4 NOTES

Section 7.1: Zwicki (1948) originated the idea of a morphological chart. Further discussion and examples of morph charts can be found in Cross (1994), Jones (1992), and Hubke (1988).

Section 7.2: The address of the USPO's website is www.uspto.gov. Another often-used website is www.ibm.com/patents. Group methods of idea generation are explored and described in Shah (1998). Approaches to creativity and analogical thinking in a group setting are described in Hays (1992).

Section 7.3: The results for the Danbury arm support design project are taken from final reports by Attarian et al. (2007) and Best et al. (2007).