Optimization of Prediction Error in CO2 Laser Cutting process by Taguchi Artificial Neural Network Hybrid with Genetic algorithm

Simulation and prediction of CO2 laser cutting of Perspex glass has been done by feed forward back propagation Artificial Neural Network (ANN). Experimental data of Taguchi orthogonal array L9 was used to train the ANN model. The simulation results were evaluated and verified with the experiment. In some cases, the prediction errors of Taguchi ANN model was larger than 10% even with Levenberg Marquardt training algorithm. To overcome such problem, a hybrid genetic algorithm-based Taguchi ANN (GA-Taguchi ANN) has been developed. The potential of genetic algorithm in optimization was utilized in the proposed hybrid model to minimize the error prediction for regions of cutting conditions away from the Taguchi based factor level points. The hybrid model was constructed in such a way to realize mutual input output between ANN and GA. The simulation results showed that the developed GA-Taguchi ANN model could reduce the maximum prediction error below 10%. The model has significant benefits in application to fabrication processes