Robust Design as a Design Philosophy

We can think more broadly about the philosophy of robust design, as compared to a traditional approach. Generally, variation exists in products, both due to the user's environment and due to manufacturing. The issue is over how to consider the variation. The standard approach is to reduce the noise space to a set of fixed values, and then select the design configuration using these fixed numbers. The noise variable space is reduced to a single point-a single value for each variable, called a constant. This approach applies even in physical prototyping, where one might use one typical set of operating and use conditions.

They are held constant. Of course, the constants are never really constant. To compensate for this reality, a factor of safety is applied to the constant, in the hopes that one can ensure the risk is acceptable. A robust design approach is different. Rather than first reducing the noise space to a constant and then making design choices, the noise space is considered, and a suite of performances are evaluated. The results are subsequently reduced to a single evaluation, but only after the range of output is understood.

The differences in philosophies are diagrammed in Figure 19.11. Notice the use of factors of safety versus norms and multiple performance evaluations.

Given these observations, it should become clear that one misses the point in debates over topics such as what norm ought to used, whether a main effect model is appropriate, or what Taguchi's logarithmic transformation really does. The primary concern that makes robust design effective is that the noise space should be considered and reduced only on the performance space. Reducing the noise space variations before any evaluations are made is only safe when the mapping is well understood-we believe we know what will happen with the introduction of the noise.

The issue, of course, reduces to one of comparing the cost of repeated performance evaluations versus the confidence one has in not doing so. Performing the many more evaluations required by robust design will generally be more costly during product development: physical experiments, finite element analysis, or computer calculations are required to evaluate any performance variable value, and these generally are not trivial in time or expense.

Making the decision over when to take a robust design approach is one of engineering and business judgement. Generally speaking, though, it is almost always the case that a design team is better off in the robust design approach, even in the early conceptual phases of design. It is better to spend the extra resources during product development. This investment will ensure that there is less opportunity for failure at later stages due to unforeseen performance responses due to noise .

Golden Nuggets

Robust design is a philosophy in design as well as a set of methods. When implemented, it will slow the design process down, but will increase the effectiveness and improve the result. Some key ideas in robust design include:

• Always consider the impact that input variation will have upon the delivered performance.
• Repeated evaluations over the noise variations are a much more accurate approach than choosing a nominal value of noise and estimating with a factor of safety.
• Taguchi's method is a simple experimental approach to robust design that is very effective.
• Variations are usually the study of random noise, and expectation integration forms the basis of modeling variation.
• Worst case norms are better suited for constraint type performance variables that must be ensured. Euclidean norms are better suited for objective function type performance variables that are always better when smaller.

Date: 2016-03-03; view: 780