Response Surface Method and Neural Network to Determine Surface Roughness for Laser Cutting on Acrylic Sheets

This paper presents the use of response surface method (RSM) and neural network to study surface roughness for laser beam cutting on acrylic sheets. Box-Behnken design based on response surface method and multilayer perceptions neural network were used to predict the effect of laser cutting parameters. These parameters include power requirement, cutting speed and tips distance on surface roughness during the machining of acrylic sheets. It is found out that the predictive models are able to predict the longitudinal component of the surface roughness close to those readings recorded experimentally with a 95% confident interval. The result obtained from the predictive model was also compared using multilayer perceptions with back–propagation learning rule artificial neural network. The first order equation revealed that power requirement was the dominant factor which was followed by tip distance, and cutting speed. The cutting parameter predicted by using neural network was in good agreement with that obtained by RSM. This observation indicates the potential of using response surface method in predicting cutting parameters thus eliminating the need for exhaustive cutting experiments to obtain the optimum cutting condition to enhance the surface roughness