Forty-four rules were used to relate the seven inputs and
four outputs. Weight factors between zero and one were
assigned to each rule in order to reflect the different levels
of influence the four process parameters have on the
process metrics. The weight factors were assigned based
on survey data and also based on parametric data available
from previous experiments.
The model calculations were performed using a Mamdani
structure [3], with min and mux operators for fuzzy
intersection and union respectively, minimum implication
[4], and centroid defuzzification [2] scaled by the rule
weighting factors discussed above. The fuzzy inference
system components were designed using the MATLAl3
Fuzzy Logic Toolbox software on a personal computer.