An abstraction of the Taguchi principles is used as the basis for mathematical formulation of quality modeling in laser welding. Another abstraction is proposed that measures the discrepancy between the model and target function as the squared measure. Quality modeling is based on an artificial intelligence algorithm that transforms the quantitative into qualitative aspects of the process input and target information. So called n-graphs or qualitative 2D graphs are formed and the modeling is performed according to qualitative similarity of the model and target graphs. The measure of discrepancy and similarity between these graphs is given as an exact measure. The obtained results show that, for an expert chosen set of input variables and parameter settings, a fast and precise enough nonlinear model of the laser quality welding can be obtained.

Fizika,Elektrotehnika