Statistical Experimental Design
The objective of statistical experimentation is to determine the relationship between a dependent response variable
(the quantity observed or measured for a particular combination of factor levels) and the factors (the processing conditions which
are controlled for the experiment). As applied in this study of the pultrusion process, the dependent response variables may be a
desired property such as shear strength and the independent factors are process parameters such as pull speed or die
temperatures; however, statistical experimental design techniques can be applied to any process in which factors affect the outcome
of the process. Statistical methods of experimentation provide the ability to draw conclusions that would not be possible if traditional
one-factor-at-a-time methods of experimentation were used. For example, a central composite design consisting of thirty-two experiments
can provide more information about the effects of five factors at five levels than hundreds of experiments that changed one factor at a time.
Many pultrusion experiments conducted by the authors to determine the effects of process parameters on mechanical properties have used
a five factor, half-factorial CCD that required thirty-two experiments. A typical design matrix includes information concerning the pultrusion
processing parameters (factors) examined and levels of each factor. The process levels are normally held within ranges expected for
commercial pultrusion of the resin system being examined. All pultrusion studies are conducted in the Composite Materials Laboratory
at the University of Mississippi using a Pulstar 804 pultruder with hydraulic, reciprocating clamp pullers. A data acquisition system
connected to a personal computer is used to record set line speed, actual line speed, pull force, puller system pressure, clamp
pressure, and platen temperatures at five second intervals during each experiment. A chart recorder is also used to continuously
graph pull force during each experiment. This data collection allows the properties of the pultruded composites to be correlated
with the manufacturing data.
