(1994). Mathematical modelling of weld phenomena. International Materials Reviews: Vol. 39, No. 1, pp. 46-47.

Mathematical modelling of weld phenomena

This study proposes a method for determining the near-optimal settings of welding process parameters using a controlled random search (CRS) wherein the near-optimal settings of the welding process parameters are determined through experiments. The method suggested in this study is used to determine the welding process parameters by which the desired weld bead geometry is formed in gas metal arc (GMA) welding. In this method, the output variables (front bead height, back bead width, and penetration) are determined by the input variables (wire feed rate, welding voltage, and welding speed). The number of levels for each input variable and the total search points were determined to be 10 and 1000, respectively.

Determination of optimal welding conditions with a controlled random search procedure

A statistical process design method is proposed for robotic gas metal arc (GMA) welding of sheet metal. The proposed method has the objective of minimizing weld cycle time by maximizing welding travel speed, while maintaining predictable weld quality over a range of worst-case processing conditions. This formulation offers a coherent alternative to Taguchi signal-to-noise ratios that permits the use of classically designed experiments that have potential advantages in experimental cost savings. The method is based on recommended choices of independent variables that place all experimental points in the region of interest. Also, so relatively simple weld evaluation approach was included for sheet metal applications. The method was applied on 16-gauge, 409 stainless steel lap joint welds in the horizontal position. The optimization revealed interesting dependencies of achievable travel speeds on root opening and welding gun positioning factors.

Statistical process design for robotic GMA welding of sheet metal

Disturbances and changes in the welding working environment lead to possible variations in the output variables associated with weld quality. In order to enhance weld quality, it is essential to optimize the welding process by taking the variance as well as the average value of the output variables into consideration. In this study, the dual response approach is adopted to determine the welding process parameters, which produce the target value with minimal variance. The dual response approach optimizes the penetration in gas metal arc (GMA) welding through the procedures as follows. First, the regression models of the mean value and standard deviation of the penetration are induced through regression analysis. Subsequently, an optimization algorithm based on the regression models and constraints is applied to determine the welding process parameters, which generate the desired penetration with minimized variance. In particular, the genetic algorithm, a global optimization algorithm, is adopted in this st...

Optimization of GMA welding process using the dual response approach

Decision making in welding is very important for achieving a good quality welded joint for the least possible cost. Of particular interest is decision making involving the selection of process, parameters, weld procedure specification, defect analysis and trouble shooting. This research has provided a means of capturing the planning knowledge in a Neuro-Expert System in a form that is capable of teaming new information, correcting old information and automating the decision-making process in a welding environment. A strategy is formulated for the representation of knowledge in the form of a neural links and the translation of rules into neural link weights. After training those weights were converted back into rules to find out the inconsistent rules and capture new rules using a new approach. The various job variables affecting the process of welding are identified in detail and a Neuro-Expert system for the selection of process, parameters and weld procedure specification is developed. The neural networks are integrated with an expert system for decision making in welding environment. Apart from providing the initial parameters of welding, the expert system is used to validate the output of the neural network and served as a user-friendly interface for the neural network. Defect Analysis is performed in welding domain by mapping the welding parameters and defect patterns in a neural network. A neural network based approach for representing the knowledge in expert system is utilized for this purpose as the modification and updating of the knowledge was easier. xi Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

A Neuro-Expert Approach for Decision -Making in Welding Environment.

Statistical Process Design (SPD) has been evaluated as a method of establishing which factors of an auto- mated flux cored arc welding (FCAW) process are most critical to minimize weld defect incidence. In this specific case, eighteen tests were conducted in a SPD program to establish the critical factors. Using conventional techniques, more than 2,000 tests would have had to been conducted. Once the critical factors had been established, comparative welding tests were conducted to validate the SPD results. For this situation, the weld defect incidence was reduced from 47% to 0.074%, proving the SPD technique could be applied to this specific welding prac- tice, and it had the potential to be used on other automated welding processes.

I. A. Mcconnell

Papers

The Application of Statistical Process Design to a FCAW Process