Master of Science in Industrial Engineering (MSIE)
This thesis develops a computer-aided weld inspection methodology based on fuzzy modeling with selected features. The proposed methodology employs several filter feature selection methods for selecting input variables and then builds fuzzy models with the selected features. Our fuzzy modeling method is based on a fuzzy c-means (FCM) variant for the generation of fuzzy terms sets. The implemented FCM variant differs from the original FCM method in two aspects: (1) the two end terms take the maximum and minimum domain values as their centers, and (2) all fuzzy terms are forced to be convex. The optimal number of terms and the optimal shape of the membership function associated with each term are determined based on the mean squared error criterion. The fuzzy model serves as the rule base of a fuzzy reasoning based expert system implemented. In this implementation, first the fuzzy rules are extracted from feature data one feature at a time based on the FCM variant. The total number of fuzzy rules is the product of the fuzzy terms for each feature. The performances of these fuzzy sets are then tested with unseen data in terms of accuracy rates and computational time. To evaluate the goodness of each selected feature subset, the selected combination is used as an input for the proposed fuzzy model. The accuracy of each selected feature subset along with the average error of the selected filter technique is reported. For comparison, the results of all possible combinations of the specified set of feature subsets are also obtained.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Ghazavi, Sean Najm, "Computer-aided weld inspection by fuzzy modeling with selected features" (2007). LSU Master's Theses. 452.
T. W. Liao