Optimization of ANN models using different optimization methods for improving CO2 laser cut quality characteristics

Determination of optimal laser cutting parameter settings for obtaining high cut quality in CO2 laser cutting process is of great importance. In this paper an attempt has been made to apply different optimization methods for determining of optimal values of laser power, cutting speed, assist gas pressure and focus position with the purpose of improving the cut quality characteristics obtained in the CO2 laser cutting of stainless steel. The laser cutting experiment was planned and conducted according to the Taguchi’s L27 orthogonal array and the experimental data were used for developing mathematical models for surface roughness, kerf width and width of heat affected zone based on artificial neural networks (ANNs). Mathematical models of the cut quality characteristics were developed using single hidden layer ANN trained with Levenberg–Marquardt algorithm. This paper compares the quality of solutions obtained when optimizing ANN models using the real coded genetic algorithm (RCGA), simulated annealing (SA) and recently developed improved harmony search algorithm (IHSA). The computer code was written in MATLAB to integrate the ANN-based process models and the RCGA, SA and IHSA algorithms. For the purpose of comparison, some performance criteria were used. The merits and the limitations of the selected optimization methods were discussed.