Axiomatic designs

An approach that seeks to establish a scientific basis for design and create a theoretical foundation and associated tools, is the methodology developed by the MIT professor Nam Suh. Although his technique, known as axiomatic design, is not usually considered a quality methodology, proper application of its tools will result in robust products capable of exceeding customer expectations. It is also well suited to software engineering. The fundamental concepts of this technique are summarized in two axioms:

• Maintain the independence of functional requirements (independence axiom).

• Minimize the information content of the design (information axiom).

There are four domains within axiomatic design:

• Customer domain (attributes).

• Functional domain (functional requirements and constraints).

• Physical domain (design parameters).

• Process domain (process variables).

Items identified within one domain are mapped to the next. This facilitates the development of a design matrix that relates the functional requirements to the design parameters. Once the matrix is developed, successful application of the two fundamental axioms and related corollaries and theorems guides engineers to produce designs that avoid coupling.

In coupled designs, one design parameter influences multiple functional requirements. Therefore, those parameters cannot be optimized for one requirement without compromising another. For example, a water faucet with separate hot and cold valves is a coupled design—adjusting one valve changes both the water flow rate and its temperature. A faucet with one control lever is uncoupled— flow and temperature are controlled independently.

Successful execution of axiomatic design's algorithmic approach enables engineers to avoid the pitfalls of coupled designs and develop robust products.

In conjunction with the axiomatic design, we also may want to use signal flow graphs. These are graphs that represent relationships among a number of variables. A special condition exists when these relationships are linear. When that happens, the graph represents a system of simultaneous linear algebraic equations. For more on these topics see Eppinger, Nukala and Whitney (1997); Stamatis (2003), and Suh (1990).