Modeling of TIG welding process using conventional regression analysis and neural network-based approaches

Abstract

Conventional regression analysis was carried out on some experimental data of a tungsten inert gas (TIG) welding process (obtained from published literature), to find its input–output relationships. One thousand training data for neural networks were created at random, by varying the input variables within their respective ranges and the responses were calculated for each combination of input variables by using the response equations obtained through the above conventional regression analysis. The performances of the conventional regression analysis approach, a back-propagation neural network (BPNN) and a genetic-neural system (GA-NN) were compared on some randomly generated test cases (experimental), which are different from the training cases. It is interesting to note that for the said test cases, the NN-based approaches could yield predictions that are more adaptive in nature compared to those of the more conventional regression analysis approach. It could be due to the fact that NN-based approaches are able to bring adaptability, which is missing in the conventional regression analysis. Moreover, GA-NN was found to perform better than the BPNN, in most of the test cases. A BPNN works based on the principle of a steepest descent method, whose solutions have the chance of being trapped at the local minima, whereas in GA-NN, the search for a minimum deviation in prediction, is carried out using a GA. However, their performance depends on the nature of the deviation function.

Introduction

To ensure both high productivity as well as good quality of the products, a manufacturing process is to be automated. In order to automate a process, a proper model has to be constructed and tested before implementing for on-line control. This paper deals with modeling of a tungsten inert gas (TIG) welding process. There is a natural quest of the researchers to establish input–output relationships of a process. Rosenthal [1] studied the temperature distributions on an infinite sheet, due to a moving point heat source considering the heat dissipation by conduction. His analysis could be related to arc welding after making a number of assumptions. However, he never tried to relate his theoretical solution to the weld bead geometry, which was attempted later on by Roberts and Wells [2]. Later on, a considerable amount of work have been carried out on analytical modeling of welding process by various investigators. In this connection, the work of Bhadeshia [3] is worth mentioning. A model was developed by Bhadeshia et al. [4] to study the process of micro-structure formation in low-alloy steel weld deposits. Svensson et al. [5] carried out an analysis of cooling curves for the fusion zone of steel weld deposits. The cooling curves were obtained for a wide range of welding current, voltage, speed and inter-pass temperature. Moreover, Bhadeshia [6] developed the model of phase transformations and micro-structure formation in steel welds. Two-dimensional axi-symmetric finite element analysis of conduction heat flow in laser spot formation was done by De et al. [7]. They could also predict the cooling rate and micro-structure formation in laser spot welds [8]. It might be difficult to model a complicated process like welding analytically. Realizing this fact, several attempts were made by various investigators to model the welding process by using some conventional regression analysis approaches. Both the linear as well as non-linear conventional regression analyzes had been carried out in the past, based on the experimental data collected in a particular fashion (e.g., full factorial design of experiments, fractional factorial design of experiments). Some of these works are mentioned below. Yang et al. [9] used a non-linear regression analysis for modeling a submerged arc welding process. Murugan et al. [10] utilized a response surface methodology to establish the relationships between the input process parameters and bead geometric parameters, in case of a submerged arc welding process. Tarng and Yang [11] applied the Taguchi method, to carry out optimization of weld bead geometry in a gas tungsten arc welding. Lee and Rhee [12] conducted multiple regression analysis to perform both forward (i.e., from process parameters to weld bead geometric parameters) as well as backward (i.e., from bead geometric parameters to process parameters) mappings, in case of a gas metal arc welding process. Kim et al. [13] derived both linear as well as non-linear multiple regression equations to establish the relationships between the process variables and bead penetration for a robotic CO₂ arc welding process.

Several trials had been made by various researchers to model the welding processes using artificial neural networks (ANNs). Some of their works are discussed below. Anderson et al. [14] used the ANN to model an arc welding process and concluded that it can provide an accuracy in prediction comparable to that of other conventional control systems. Cook et al. [15] successfully developed an ANN model to investigate three areas of welding analysis, viz. process modeling, control and quality of weld beads. To relate input process parameters and weld bead geometric parameters in TIG welding, an attempt was made by Juang et al. [16] by using both the back-propagation as well as counter-propagation networks. They concluded that the back-propagation algorithm has better learning ability, whereas the counter-propagation algorithm has better generalization ability. Wu et al. [17] developed a self-organizing map (SOM), for monitoring and quality evaluation in a gas metal arc welding process. More recently, Nagesh and Datta [18] developed a back-propagation neural network, to establish the relationships between the process parameters and weld bead geometric parameters, in a shielded metal arc welding process. Back-propagation neural network (BPNN) had been widely used to model the welding processes. But, the chance of its solutions getting trapped into the local minima is high, as it works based on the principle of a steepest descent method. To overcome this difficulty, a genetic algorithm (GA) [19] may be utilized (in place of the steepest descent method), along with the feed forward NN, which may be termed as a genetic-neural system (GA-NN).

In the present paper, a TIG welding process has been modeled by using a conventional linear regression technique, a BPNN and a GA-NN and their performances are compared using some test cases. The rest of the text has been organized as follows. Section 2 introduces the bead geometric parameters obtained in a TIG welding process. Section 3 explains the modeling schemes using three different approaches—conventional regression analysis, back-propagation neural network and genetic-neural system. Results are discussed in Section 4. Some conclusions are made in Section 5.