Prediction and optimization of weld bead geometry for electron beam welding of AISI 304 stainless steel

Abstract

Electron beam welding, though considered a sophisticated welding process, still requires the operator to first carry out several trial welds to find the right combination of welding parameters based on intuition and experience. This archaic method is often unreliable, leading to unproductive manufacturing lead time, man hours, quality control tests, and material wastage. The current study eliminates this "trial and error" method by providing a reliable model which can predict the right combination of weld parameters to achieve a high-quality weld. Beads on plate welds were carried out on AISI 304 stainless steel plates using a low-kilovolt electron beam welding (EBW) machine. A model that can predict weld bead geometry and provide optimized output for minimum weld area condition without compromising on weld quality was developed. Experimental data were collected as per full factorial design of experiments, and the levels for each input parameter were established through pilot experiments. A multivariate regression analysis has been conducted to establish a relationship between four weld input parameters (three levels each) and four weld bead responses. Response surface methodology (RSM) has been used to study the interrelationship between input parameters and their effect on each response variable. Further, minimization of weld cross-sectional area was done using genetic algorithm for maximum penetration and minimum weld area condition. The optimized mathematical model convincingly establishes that the focusing current is a significant input parameter with very high influence over the weld bead geometry. Extensive material characterization and mechanical tests have been carried out to validate the regressed input-output relationship and the optimized mathematical model. Keywords Electron beam welding . Weld bead geometry . Stainless steel . Full factorial design of experiment . Multiple regression analysis . Response surface methodology Abbreviations ANOVA Analysis of variance BBW Back bead width BH Bead height BP Bead penetration BW Bead width D Standoff distance/work to chamber top distance (mm) EBW Electron beam welding F Focusing current (A) HAZ Heat-affected zone I Beam current (mA) RSM Response surface methodology R-sq, R 2 Coefficient of correlation S Welding speed/beam spot travel (m/min) SE Standard error T Thickness of plate (mm) V Accelerating voltage (kV) * Arpith Siddaiah

Prediction of Bead Geometry with Changing Welding Speed Using Artificial Neural Network

“…Improving the manufacturing quality and assembly quality heavily raises the production cost, and the traditional “trial and error” method needs plenty of man hours. As such, an acceptable solution is the introduction of intelligent welding technology to improve the reliability of product and production efficiency [ 3 , 4 ]. For a given task, the intelligent welding system should eliminate the trial time substantially to specify welding parameters.…”

“…Because to the complex interrelationship between these welding parameters, it is hard to deduce an appropriate physics model in continuous welding process with changing parameters [ 5 ]. In recent decades, researchers have applied various mathematical models to build the relationship between multi-input and multi-output parameters, e.g., factorial design, linear and nonlinear regression, response surface methodology, and artificial neural network (ANN) [ 4 , 6 , 7 , 8 , 9 , 10 , 11 ]. These design of experiments (DOE) techniques apply to different areas according to the complex relationship between input and output parameters, and they achieve high accuracy and efficiency in modeling.…”

Dynamic Modeling of Weld Bead Geometry Features in Thick Plate GMAW Based on Machine Vision and Learning

“…Weld bead geometry features (WBGFs) can reflect the welding process parameters and are effective evidence to suggest how to optimize the latter. This is because the relationship between WBGFs and weld process parameters can be built with various models [ 9 , 10 , 11 ]. Thus, the real-time modeling of WBGFs is the really demanded technology for the effective control of weld formation especially and accurate metal deposition [ 12 , 13 ] during the multipass welding process.…”

A Virtual Design of Experiments Method to Evaluate the Effect of Design and Welding Parameters on Weld Quality in Aerospace Applications

“…In the literature, studies can be found that use design of experiments (DOE) and other statistical and numerical approaches to predict and optimize weld quality for different welding methods [14,[21][22][23][24][25][26][27][28]. Nevertheless, few studies can be found that optimize the interaction of product geometry and welding process.…”