Application of RCGA-ANN approach for modeling kerf width and surface roughness in CO2laser cutting of mild steel

This paper presents an artificial intelligence approach for the development of predictive models for a CO2laser cutting of mild steel by using artificial neural networks (ANNs) and real coded genetic algorithm (RCGA). Laser cutting experiment, conducted according to Taguchi’s experimental design using L25orthogonal array, provided a set of data for the development of ANN models for the prediction of the kerf width and surface roughness. Both ANN models considered cutting speed, laser power, and assist gas pressure as input parameters. Considering the disadvantages of the back propagation, the RCGA was applied for training of the ANNs. Statistical results indicate good correlation between the experimental results and ANN predictions, which confirms the validity of the applied approach. Finally, using the developed models, the combined effects of input process parameters on the quality characteristics were studied.