Information Technology Reference
In-Depth Information
experiment (Anderson, 2000). More importantly, the estimates of the factor effects
from OFAT are less precise than DOE. In many situations, the effects of a factor
would change when the conditions of other factors vary. Using DOE, one can vary
all the factors simultaneously. This allows the experimenter to determine the effect
of a factor when the levels of the other factors change. OFAT techniques do not
determine the interactions among the different factors consequently leading to inac-
curate test results. DOE provides better product or process optimization tools than
traditional OFAT experimentation. This applies to experiments where the response
function has to be maximized or minimized. A regression model can be created
which highlights the relationship between key input factors and factor interactions,
something which OFAT is unable to provide. Regression models can also provide
engineers with extremely important or valuable information; for example, by study-
ing the model the setting of factors can be manipulated to achieve a pre-determined
target level for the variable of interest.
As mentioned previously, DOE is one of the most powerful quality improvement
techniques for reducing process variation, enhancing process effectiveness and pro-
cess capability. If implemented well, DOE can optimize the process or product under
investigation “by exploiting the non-linear effects of the process parameters on the
performance characteristics” (Simms & Garvin, 2002) and help determine the set-
tings which would minimize variation within that product or process. However, what
is the cause of this variation?
Common variation within a process is known as “noise.” This noise can occur
as a result of poorly maintained machinery, inconsistent operating conditions from
operator skill levels, machine fluctuations, same batch raw material variation, etc.
Optimizing a process through DOE involves determining the optimal setting condi-
tions of each factor which would lead to a response which is least sensitive to this
noise. If all the control and signal factors are optimized, the result is a robust process
with significantly reduced non-conformance and variability. Taguchi and Clausing
(1990) stated that “if a process performs well in adverse conditions (as a result of
noise factors) it will perform considerably better in normal conditions.” As the cor-
rect implementation of DOE is highly critical for the success of an experiment, a
number of distinct steps following the plan-do-check-act procedure have been put
forward (Antony &Knowles, 2001; Montgomery, 2001b; Simms and Garvin, 2002).
Unfortunately, they all differ slightly and so the steps to completing a DOE are not as
readily recognized. However, used within the context of a structured methodology,
DOE approach such as RDM can lead to beneficial outcomes.
Since the 1980s engineers have gradually become more aware of the benefits of
using DOE and as a result for this there has been many new opportunities for apply-
ing DOE. The most important of these is RDM - a methodology that was developed
by Taguchi (Arvidsson & Gremyr, 2008). Its advantage over other techniques is
that it enables robustness due to its emphasis on designing processes insensitive to
uncontrollable factors known as noise.
RDM has been defined as “ Systematic efforts to achieve insensitivity to noise
factors. These efforts are founded on an awareness of variation and can be applied
in all stages of product design. ” (Arvidsson & Gremyr, 2008).
Search WWH ::




Custom Search