Gmaw bead geometry prediction using regression analysis and ann

GMAW BEAD GEOMETRY PREDICTION USING REGRESSION ANALYSIS AND ANN D. Ramos-Jaime, I. Lopez-Juarez, U Kruger Centro de Investigación y Estudios Avanzados del IPN-Unidad Saltillo. The Gas Metal Arc Welding (GMAW) is a very complex process, which is perhaps one of the industrial processes more difficult to model and control since four states of different materials coexist in a small volumesimultaneously (solid, gaseous, liquid and plasma) and where interactions associated with the electrical, magnetic, kinetic, thermal, chemical, atomic processes and fluidic occur apart from the ones associated with the automation itself (e.g. robotic arc welding). In this work, we study the geometrical properties of the bead weld namely: width, height and penetration, which are the most important parametersto asses, the quality of the weld. We designed a 23 experiment to identify the most appropriate values of arc voltage, wire feed speed (associated to current) and torch travelling speed. From the resulting data we develop a regression analysis and compared it with the obtained with the Backpropagation Artificial Neural Network whose results are presented in this paper. Keywords: Arc welding,Regression analysis, Design of experiments, Industrial robots. the process parameters used in a full factorial design, for this they use three input variables (voltage, current and speed of travel) and three output variables (height, width and depth of the weld). Kumar and Debroy (2004) performed a model based on equations of heat transfer involved in the process and performs and optimization of theparameters using two conjugate gradient methods, as well as the method of Levenberg–Marquardt. In this paper we present a fractional factorial experimental design, compare its results using linear regression and the Backpropagation algorithm. 2. EXPERIMENTAL SYSTEM The article presents the developed experimental system. This system consist of a KUKA KR16 industrial robot, 455M Lincoln power sourceand modified gas and wire feeders. The control system consists of a PC master computer, which provides communication at two levels, thus positioning the robot’s end effector via serial port using the 3964R protocol as well as communication with a data acquisition card that controls the onset of electric arc and gas injection. The system is controlled by an application program developed in C++,which was validated experimentally using A1018 steel plates and whose geometric parameters of the bead were evaluated in the metallographic laboratory.

1. INTRODUCTION In the last few years automated welding systems have received a great deal of attention because they are highly suitable, not only for increasing the production and quality, but also to decrease cost and time of manufacture for agiven product. Of these systems, the robotic GMAW (Gas Metal Arc Welding) is one of the most easily found in any industry whose products requires metal joining in a large scale. It establishes an electric arc between a continuous filler metal electrode and the weld pool; this is shielded by an externally supplied gas. The heat of the arc melts the surface of the base metal and the end of theelectrode. The electrode molten metal is transferred through the arc to the work piece where it becomes the deposited weld metal (weld bead). (Sampaio et al, 2004). The GMA welding parameters that are most important factors affecting the quality, productivity and cost of welding joint are the weld bead size and shape. On an industrial scale the automation of this process is performed by servomechanismsand retention mechanisms such as special fixtures, belts and anthropomorphic robots adding uncertainty parameters to process such as positioning, ageing components, mechanisms, etc., which together with the physical phenomenon affect the quality of the welding and whose evaluation is carried out primarily by the geometry of the cord welding (width, height, and penetration) errors. 2. RELATED WORK...