A robust product is one that works as intended regardless of variation
in a product's manufacturing process, variation resulting from deterioration,
and variation in use. Robust design can be achieved when the designer understands
these potential sources of variation and takes steps to desensitize the
rpoduct to these potential sources of variation. Robust design can be achieved
through "brute force" techniques of added design margin or tighter
tolerances or through "intelligent design" by understanding which
product and process design parameters are critical to the achievement of
a performance characteristic and what are the optimum values to both achieve
the performance characteristic and minimize variation.
When the operation of the product or achievement of a performance characteristic
can be mathmatically related to a product or process design parameter, optimum
product and process design parameters can be calculated. When these relationships
are unknown, design of experiments (DOE) can aid in determining these optimum
parameter values and, thereby, developing a more robust design.
Design of Experiments
Design of Experiments is based on the objective of desensitizing a product's
performance characteristic(s) to variation in critical product and process
design parameters. Genichi Taguchi developed the concept of "loss to
society". In this concept, variability in critical design parameters
will increase the loss to society which is an expanded view of the traditional,
internally-oriented cost of quality. This is a quadratic relationship of
increasing costs (loss to society) as these critical design parameter values
vary from the desired mean value of the parameter.
To consider quality implications during design, the design process can be
segmented into three stages. The first stage, system design, establishes
the functionality of the product, the physical product envelope, and general
specifications. The second stage, parameter design, establishes specific
values for design parameters related to physical and functional specifications.
It is during these first two stages that the designer has the greatest opportunity
to reduce product costs through effective functional design and parameter
specification. The third stage, tolerance design, establishes the acceptable
tolerances around each parameter or target. The third stage typically will
add costs to the product through efforts to ensure compliance with the tolerances
associated with product parameters.
Since an organization cannot cost-effectively inspect quality into the product,
it must focus on minimizing variability in the product through product and
process design and control of processes. However, some variability is uncontrollable
or very difficult to control. This difficult to control variation is referred
to as noise. Noise is the result of variation in materials, processes, the
environment and the product's use or misuse. Products need to be designed
so that they are robust - their performance is insensitive to this naturally
occurring, difficult to control variation.
Design of Experiments techniques provide an approach to efficiently designing
industrial experiments which will improve the understanding of the relationship
between product and process parameters and the desired performance characteristic.
This efficient design of experiments is based on a fractional factorial
experiment which allows an experiment to be conducted with only a fraction
of all the possible experimental combinations of parameter values. Orthogonal
arrays are used to aid in the design of an experiment. The orthogonal array
will specify the test cases to conduct the experiment. Frequently, two orthogonal
arrays are used: a design factor matrix and a noise factor matrix, the latter
used to conduct the experiment is the presence of difficult to control variation
so as to develop a robust design. This approach to designing and conducting
an experiment to determine the effect of design factors (parameters) and
noise factors on a performance characteristic is represented below.
These experimental results can be summarized into a metric called the
signal to noise ratio which jointly considers how effectively the mean value
(signal) of the parameter has been achieved and the amount of variability
that has been experienced. As a result, a designer can identify the parameters
that will have the greatest effect on the achievement of a product's performance
The design parameters or factors of concern are identified in an inner array
or design factor matrix which specifies the factor level or design parameter
test cases. The outer array or noise factor matrix specifies the noise factor
or the range of variation the product will be exposed to in the manufacturing
process, the environment or how the product used (conditions it is exposed
to). This experimental set-up allows the identification of the design parameter
values or factor levels that will produce the best performing, most reliable,
or most satisfactory product over the expected range of noise factors or
After the experiments are conducted and the signal to noise ratio determined
for each design factor test case, a mean signal to noise ratio value is
calculated for each design factor level or value. This data is statistically
analyzed using analysis of variation (ANOVA) techniques. Very simply, a
design factor with a large difference in the signal noise ratio from one
factor setting to another indicates that the factor or design parameter
is a significant contributor to the achievement of the performance characteristic.
When there is little difference in the signal to noise ratio from one factor
setting to another, this indicates that the factor is insignificant with
respect to the performance characteristic.
With the resulting understanding from the experiments and subsequent analysis,
the designer can:
- Identify parameter values which maximize achievement of performance
characteristic and minimize the effect of noise, thereby achieving a more
- Identify parameters that have no significant effect on performance.
In these cases, tolerances can be relaxed and cost reduced.
- Identify parameter values which reduce cost
without affecting performance or variation.
These steps take initial effort, but can reduce cost and improve the performance
of the product. In the past, the designer selected design parameters and
tolerances and made system design trade-offs in an intuitive manner, sometimes
supported by limited analysis and trial and error experimentation. However,
an overall framework was lacking to make these decisions. Design of Experiments
techniques offer a framework for developing a more rigorous understanding
of the relationship between product and process parameters and the achievement
of a performance, reliability or quality characteristic, thereby leading
to improved design decisions. These techniques present a comprehensive approach
experimental design, analysis, and product and process design decision-making.
ABOUT THE AUTHOR
Kenneth A. Crow is President of DRM Associates,
a management consulting and education firm focusing on integrated product
development practices. He is a distinguished speaker and recognized expert
in the field of integrated product development. He has over twenty years
of experience consulting with major companies internationally in aerospace,
capital equipment, defense, high technology, medical equipment, and transportation
industries. He has provided guidance to executive management in formulating
a integrated product development program and reengineering the development
process as well as assisted product development teams applying IPD to specific
He has written papers, contributed to books, and given many presentations
and seminars for professional associations, conferences, and manufacturing
clients on integrated product development, design for manufacturability,
design to cost, product development teams, QFD, and team building. Among
many professional affiliations, he is past President and currently on the
Board of the Society of Concurrent Engineering and is a member of the Product
Development Management Association and the Engineering Management Society.
For further information, contact the author at DRM Associates, 2613 Via
Olivera, Palos Verdes, CA 90274, telephone (310) 377-5569, fax (310) 377-1315,
or email at email@example.com.