الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 264 |  |  |
| | | from | 264 | | |
Abstract
Laser beam cutting is one of the major applications of lasers in sheet metal working. In this thesis, an experimental study in CO2 laser cutting process is presented. The aim of this research is to investigate the effect of the laser cutting process variables on the cutting-edge quality parameters. Further to develop mathematical models and optimize the various laser cutting variables for each of the studied materials. Two different techniques namely: statistical regression and artificial neural networks were used to develop the mathematical models. Three difficult to cut materials namely: stainless steel 316, Armox500T, and aluminum AG5 were chosen as workpiece materials in this research. A 4.4 kW CO2 Bystar L 4025-65 industrial laser cutting machine was used to perform the cutting operations. Several experiments were conducted to investigate the influence of four input variables: focal plane position, assist gas pressure, laser power, and cutting speed on the four most important performance parameters, namely: upper kerf width, lower kerf width, kerf taper angle, and the arithmetic average surface roughness Ra. The experimental plan was performed using L32 12×34 Taguchi standard orthogonal matrix. The developed ANN models are based on multilayer feed-forward neural networks. The experimentally acquired data was used to train, validate and test the ANN performance, and special graphs were drawn for this purpose. Minitab software was used to build up the mathematical models using regression analysis. As well as to generate the main effects plot for the process variables on the performance parameters in order to determine the most significant variables. The developed models were validated and proved their capability to predict the laser cutting process output parameters for certain input variables inside the covered range of this study. Genetic optimization search algorithm was used to suggest the near optimal setting combinations of the four input variables in their operating range, to achieve a minimum of surface roughness Ra and kerf taper angle θk.
This research would provide a good demonstration for the most significant input variables and new models based on regression, and ANN technique to predict the cutting-edge quality parameters, which can be used for solving related industrial problems.