100-Metodos-de-Qualidade-Total

122 100 METHODS FOR TOTAL QUALITY MANAGEMENT
Method 50
Tolerance design
Purpose
To find out by experiment where the variability in a process (product)
occurs and where adjustments can be made.
When to use
When the influences of inner and outer sources of 'noise' cannot be
successfully reduced by use of parameter design (Method 42). Here 'noise'
represents the effects of uncontrollable factors.
How to use
It requires the same steps as parameter design, but additional factors are
considered that were previously excluded because of cost or the difficulty
of experimentation.
If this also fails, the tolerance of the product's components are
considered. This means retaining the optimum nominal levels of actors,
but reducing the tolerance of certain crucial factors in an optimal and costeffective
way so that overall variability in the response is reduced to
acceptable levels.
Benefits
It assists in the study of factors which are expensive and difficult to change
in experiments for the improvement of a process.
Example
If we run a second outer array centered on the optimum parameter
conditions, this time the noise factors are noise in the parts themselves. If
factor A is a resistor at a nominal level of 100 ohms, it becomes, in the
noise array, a factor with levels of, say, 100 ± 2. When the array has been
run, we find how much variability is associated with each factor and, hence,
where we have to make further improvements.
Reference
P.W.M. John (\990) Statistical Methods in Engineering and Quality Assurance. New York:
Wiley.