Global versus cluster-wise regression analyses for prediction of bead geometry in MIG welding process

In this paper, an attempt is made to establish input-output relationships in MIG welding process through regression analyses carried out both globally (i.e., one set of response equations for the entire range of the variables) as well as cluster-wise. It is important to mention that the second approach makes use of the entropy-based fuzzy clusters. The investigation is based on the data collected through full-factorial design of experiments. Results of the above two approaches are compared and some concluding remarks are made. The cluster-wise regression analysis is found to perform a slightly better than the global approach in predicting weld bead-geometric parameters. (c) 2007 Elsevier B.V. All rights reserved.