Você está na página 1de 2

15.6.

2 Noise Factors
The input parameters that affect the quality of the product or process may be
classifi ed as design parameters and disturbance factors. The former are parameters
that can be specifi ed freely by the designer. It is the designers responsibility to
select the optimum levels of the design parameters. Disturbance factors are the
parameters that are either inherently uncontrollable or impractical to control.
Taguchi uses the term noise factors to refer to those parameters that are either
too diffi cult or too expensive to control when a product is in service or during
manufacture of its components. The noise factors can be classifi ed into four
categories:

Variational noise is the unit-to-unit variation that nominally identical products


will exhibit due to the differences in their components or their assembly.
Inner noise is the long-term change in product characteristics over time due to
deterioration and wear.
748

engineering design

Design noise is the variability introduced into the product due to the design
process. This consists mostly of the tolerance variability that practical design
limitations impose on the design.
External noise , also called outer noise, represents the disturbance factors that
produce variations in the environment in which the product operates. Examples
of external noise factors are temperature, humidity, dust, vibration, and the skill
of the operator of the product.

The Taguchi method is unusual among methods of experimental investigation in


that it places heavy emphasis on including noise factors in every experimental
design. Taguchi was the fi rst to articulate the importance of considering external
noise directly in design decisions.

15.6.3 Signal-to-Noise Ratio


W henever a series of experiments is to be carried out, it is necessary to decide
what response o r output of the experiment will be measured. Often the nature of
the experiment provides a natural response. For example, in the control chart in
Fig. 15.5, which evaluated the effectiveness of a heat-treating process for
hardening steel bearings, a natural response was the Rockwell hardness
measurement. The Taguchi method uses a special response variable called the
signal-to-noise ratio , S/N . The use of this response is somewhat controversial,

15

but its use is justifi ed on the basis that it encompasses both the mean (signal) and
the variation (noise) in one parameter, just as the quality loss function does. 1
Following are three forms of the S/N ratio corresponding to the three forms
of the loss function curves shown in Fig. 15.6.
For the nominal-is-best type of problem,
2

S / N =10 log

where = n

(15.15)

ln

2=

yi and

1 (
n 1 i=n 1 yi

=1

15

and n is the number of external noise observation combinations used for each
design parameter matrix (inner array) combination. For example, if four tests are
made to allow for noise for each combination of the control parameters, then n =
4. For the smaller-the-better type of problem,
S / N =10

log

yi2

(15.16)

chapter 15: Quality, Robust Design, and Optimization 749

For the larger-the-better type of problem, the quality performance characteristic


is continuous and nonnegative. We would like y to be as large as possible. To fi nd
the S/N, we turn this into a smaller-the-better problem by using the reciprocal of
the performance characteristic.

. Dr. Taguchi was an electrical engineer with the national telephone system of Japan, so the concept of
signal-to-noise ratio, the ratio of signal strength to unwanted interference in a communications circuit, was
very familiar to him.

Você também pode gostar