Bad design is, well, bad design. Six-Sigma, tightening tolerances, substituting one material for another—these only treat the symptoms, not the problem. Also, they may create expensive bad designs.
Axiomatic design, a theory and methodology developed at Massachusetts Institute of Technology (MIT; Cambridge, MA) 20 years ago, helps designers focus on the problems in bad designs. Says the theory’s creator, Professor Nam Suh, “The goal of axiomatic design is to make human designers more creative, reduce the random search process, minimize the iterative trial-and-error process, and determine the best design among those proposed.”
And this applies to designing all sorts of things: software, business processes, manufacturing systems, work flows, etc. What’s more, it can be used for diagnosing and improving existing designs.
What is axiomatic design?
While “MIT” and “axiomatic” might suggest some lofty academic theory, axiomatic design is well grounded in reality. It is a systematic, scientific approach to design. It guides de-signers through the process of first breaking up customer needs into functional requirements (FRs), then breaking up these requirements into design parameters (DPs), and then finally figuring out a process to produce those design parameters. In MIT-speak, axiomatic design is a decomposition process going from customer needs to FRs, to DPs, and then to process variables (PVs), thereby crossing the four domains of the design world: customer, functional, physical, and process.
The fun begins in decomposing the design. A designer first “explodes” higher-level FRs into lower-level FRs, proceeding through a hierarchy of levels until a design can be implemented. At the same time, the designer “zigzags” between pairs of design domains, such as between the functional and physical domains. Ultimately, zigzagging between “what” and “how” domains reduces the design to a set of FR, DP, and PV hierarchies.
Along the way, there are these two axioms: the independence axiom and the information axiom. (From these two axioms come a bunch of theorems that tell designers “some very simple things,” says Suh. “If designers remember these, then they can make enormous progress in the quality of their product design.”) The first axiom says that the functional requirements within a good design are independent of each other. This is the goal of the whole exercise: Identifying DPs so that “each FR can be satisfied without affecting the other FRs,” says Suh.
The second axiom says that when two or more alternative designs satisfy the first axiom, the best design is the one with the least information. That is, when a design is good, information content is zero. (That’s “information” as in the measure of one’s freedom of choice, the measure of uncertainty, which is the basis of information theory.) “Designs that satisfy the independence axiom are called uncoupled or decoupled,” explains Robert Powers, president of Axiomatic Design Software, Inc. (Boston, MA), developers of Acclaro, a software application that prompts designers through the axiomatic design process. “The difference is that in an uncoupled design, the DPs are totally independent; while in a decoupled design, at least one DP affects two or more FRs. As a result, the order of adjusting the DPs in a decoupled design is important.”
This order that Powers speaks of is shown in a design matrix that shows functional coupling between FRs and DPs at a given level of the design hierarchy. Ideally, these FRs and DPs are to be decoupled.
Axiomatic and other design methodologies
Axiomatic design is not quite the Taguchi method, which is a specific application of Robust Design. It is not quite quality function deployment (QFD). Nor, like many other quality methodologies, is it an after-the-fact approach that looks at results and then traces back to the source of those results.
Robust Design (Taguchi) and axiomatic design are the only methods that address the design itself, ensuring that the designs are good to start with. Unfortunately, while Taguchi focuses on making a part im-mune to the error in variation, it focuses on only one requirement at a time. A problem might arise when a design has to satisfy two requirements simultaneously, such as designing a car door to seal completely and close easily. In short, a coupling exists between these two functional requirements.
Taguchi method alone, we‘ve learned from an engineer at one of the automakers, can trap designers into optimizing the wrong function, optimizing a function they don’t have ownership of, or optimizing a design parameter that’s linked to many functions. Worse, by optimizing one function, designers run the probable risk of degrading other functions.
Axiomatic design avoids all that by breaking the coupling between functional requirements so that they no longer interact with one another.
QFD is similar to axiomatic design in that customer requirements are listed along the left side of a matrix and engineering requirements are lined up along the top. From this matrix, designer teams can see conflicts that need to be resolved. However, QFD is very subjective. Nor does QFD show a mathematical relationship between a functional requirement and a design parameter, which axiomatic design does.
Applying axiomatic design to care
The automotive industry is fraught with couplings between design parameters, such as in styling versus aero-dynamic/cooling requirements, styling versus crashworthiness, and in highly complex automatic transmissions designs. At one carmaker, the axiomatic methodology helps design teams optimize elements of a conceptual design before engineering creates the detailed designs. The described benefit is that it helps avoid unintended consequences in design, with the axiomatic method indicating where there are interactions between the various elements and what the optimum sequence is.
The point is that axiomatic design is said to be a step beyond Taguchi.