1.1 Introduction to design of experiments
Experiments are performed by investigators in virtually
all fields of inquiry, usually to discover something about a particular
process or system. Literally, an experiment is a test. A designed
experiment [1] is a test or series of tests in which purposeful
changes are made to the input variables of a process or system
so that the reasons for the changes in the output responses can
be observed and identified. The investigator sets several factors
in these experiments simultaneously and changes the factor settings
from experiment to experiment in a specified manner. This procedure
yields the maximum amount of information about the effect of
input variables on the output response.
Some special statistical experiments require mere
simple arithmetic calculations to yield sufficiently precise and
reliable information. Each such special design has a rational
relationship to the purpose of experimentation, the needs of the
investigator and the physical limitations of the experiments.
All such designs begin with the statement of the investigator's
objective and the identification of the factors that have the
greatest potential influence upon the response. Some common statistical
designs are:
1.2 Comparison of various methods
Among the above designs, the design of experiments
based on the regression analysis is used to find out the exact
mathematical relationship between the cause (independent input
variable) and the effect (performance parameter) provided the
independent variable and the performance parameters are quantitative
figures. When the independent variable is an attribute that is
not a measurable quantity (e.g., gender, geographic location,
plant shift, etc.,) then the regression analysis can not be used.
This poses a major limitation for generic cases where the independent
variable is not a quantitative figure.
When the intention of the investigator is to understand
the relationship between the cause and the effect rather than
just obtaining the mathematical equations relating the cause and
the effect, then the statistical approach and Taguchi methods
are best suited.
The design of experiments using statistical approach
is a simple and systematic approach by identifying various independent
factors and levels, and conducting experiments by varying one
variable at a time. In order to reduce the noise effect or error
due to the order / sequence in which the experiments are conducted,
randomization of the sequence of experiments and variables is
done.
Statistical experiments consist of several well-planned
individual experiments conducted together. The setting up of a
statistical experiment [5] involves several steps such as the
following.
The main drawback of the statistical approach is
that there are no precise guidelines for the sequence of experiments
to be conducted and the level combinations of various independent
variables for each experiment. Moreover, the number of experiments
is, in most of the cases, more than that of experiments conducted
using Taguchi method.
The design of experiments using Taguchi method is
more efficient compared to statistical methods. By choosing proper
level combinations of various independent variables, the number
of experiments is reduced considerably. At the same time, there
is no loss of any information due to reduction of number of experiments.
The method has been described in the subsequent chapters.
1.3 Areas of application
While the design of experiments can be used in many
engineering and non-engineering areas, the following are a few
engineering areas where this technique can be effectively used.
1.3.1 Process development
Many processes typically have a large number of factors
that influence the final outcome. Identification of their individual
contributions and their intricate interrelationship is essential
in the development of such processes. For example, the efficiency
of a flow solder machine meant for soldering the printed circuit
boards has several variables [2] that can be controlled. This
includes solder temperature, preheat temperature, flux type, flux
specific gravity, solder wave depth, etc.,
1.3.2 Test and development
Testing with prototypes [6] is an efficient way to
see how the concepts work when they are put into a design. Since
the experimental hardware is costly, the need to accomplish the
objectives with the least number of tests is a top priority.
1.3.3 Analysis
In the design of engineering products and processes
the analytical simulation plays an important role in transforming
a concept into the final product design. The Taguchi approach
can be utilized to arrive the best parameters for the near optimum
design configuration with the least number of analytical investigations.
Although there are several method available for optimization,
the Taguchi method is the one that treats factors at discrete
levels. This method significantly reduces computer time.