مقایسه ای در پیش بینی بازگشت دانه های حاصل از جوشکاری گازی قوس فلزی با استفاده از تجزیه و تحلیل رگرسیون چندگانه و شبکه عصبی مصنوعی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|24546||2000||10 صفحه PDF||سفارش دهید||3040 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Optics and Lasers in Engineering, Volume 34, Issue 3, September 2000, Pages 149–158
This research was done on the basis of prediction that there is a relationship between welding parameters and geometry of the back-bead in arc welding which is a gap. Multiple regression analysis and artificial neural network were used as methods for predicting the geometry of the back-bead. The multiple regression analysis and the artificial neural network were formed, and the analysis data or verification data which were used in the formation process of the multiple regression, and the training data or test data which were used in the formation process of the artificial neural network, were used to perform the prediction of the back-bead. Through this research, it was found that the error rate predicted by the artificial neural network was smaller than that predicted by the multiple regression analysis, in terms of the width and depth of the back-bead. It was also found that between the two predictions, the prediction of the width of the back-bead was superior to the prediction of the depth in both methods.
In gas metal arc welding, the weld quality is greatly affected by the welding parameters. Especially, the welding parameters are closely related to the geometry of the back-bead, a relationship which is thought to be very complicated. Repeated experiments are needed in order to determine the optimal welding conditions among welding parameters. Also, the optimal welding conditions are determined by combined factors such as the type of base metal, the welding process, and the geometry of the welded parts. Therefore, an immense amount of data is needed in order to obtain optimal welding conditions. In reality, as this amount of experimentation is impossible, a research method which can predict the geometry of the back-bead is necessary. Investigation into the relationship between the welding parameters and bead geometry began in the mid-1900s, and regression analysis was applied to welding geometry research in 1987 by Raveendra and Parmar . Chandel  suggested the correlation between the welding process parameters and bead geometry in bead-on-plate of gas metal arc welding. He confirmed that the arc current was a major parameter in determining the bead geometry. Yang et al.  used the linear, and curvilinear models to calculate the bead height from the welding process parameters in the regression equation. Also, Il-Soo Kim et al.  empirically confirmed Yang's linear, curvilinear methodology. However, there has not been any research in real welding where a gap in butt welding has been considered. Therefore, the objective of this study is to obtain the desired weld bead in real welding by predicting the geometry of the back-bead using welding parameters in gas metal arc welding. The regression analysis and the artificial neural network were used in the research. First, a system configuration was done which would predict the bead geometry, and the two prediction methods were compared and analyzed.
نتیجه گیری انگلیسی
This study was performed based on the prediction that there would be a correlation between the welding parameters and the geometry of the back-bead in gas metal arc welding, namely butt welding. The multiple regression analysis and artificial neural network were used as methods of predicting the geometry, and the following conclusions were drawn through comparison and analysis. 1. There is a linear relationship between the welding parameters and the geometry of the back-bead under optimal welding conditions, and the width and depth of the back-bead were expressed in linear equations through the multiple regression analysis. 2. By examining the relationship between the welding parameters and the back-bead through the artificial neural network, a bead geometry prediction system was made which has a smaller error rate than the multiple regression analysis. A more accurate prediction (error) can be done in terms of the width rather than the depth of the back-bead. 3. The results of a regression analysis by the stepwise method in each stage showed the formation of the back-bead geometry. The welding speed was the most important factor in the geometry of the back-bead, followed by welding current, gap and arc voltage.