177 KOLAHAN & HEIDARI: METAL ACTIVE GAS WELDING FOR GAS PIPELINES USING REGRESSION ANALYSIS Journal of Scientific & Industrial Research Vol. 69, April 2010, pp. 177-183 *Author for correspondence Tel: +98-9153114112; Fax: +98-5118763304 E-mail: kolahan@um.ac.ir Modeling and optimization of MAG welding for gas pipelines using regression analysis and simulated annealing algorithm Farhad Kolahan * and Mehdi Heidari Department of Mechanical Engineering, Ferdowsi University of Mashhad, PO Box 91775-111, Mashhad, Iran Received 16 October 2009; revised 15 January 2010; accepted 28 January 2010 This study established input-output relationships for metal active gas (MAG) welding for gas pipelines. Regression analysis (RA) was performed on data collected as per Taguchi design of experiments. Adequacy of RA model was verified using ANOVA method. RA model was then embedded into a simulated annealing (SA) algorithm to determine optimal process param- eters for weld bead geometry specification. Proposed method is found quite effective in predicting process parameters for weld bead geometry. Keywords: MAG welding, Modeling, Process parameters optimization, SA algorithm, Weld bead geometry Introduction Quality of a weld joint is directly influenced by welding input parameters during welding. A common problem faced by manufacturer is control of process input parameters to obtain a good welded joint with required bead geometry and weld quality with minimal detrimental residual stresses and distortion. Roberts & Wells 1 estimated weld bead width by considering conduction heat transfer. Christensen et al 2 derived non dimensional factors to relate bead dimensions with operating parameters. Chandel et al 3 presented predictions of effect of current, electrode polarity, electrode diameter, and electrode extension on melting rate, bead height, bead width and weld penetration in submerged arc welding. Markelj & Tusek 4 mathematically modeled current and voltage in TIG welding. Kim et al 5 conducted a sensitivity analysis of a robotic GMAW (gas metal arc welding) process employing non-linear multiple regression analysis (RA) for modeling process and quantified respective effects of process parameters on weld bead geometry (WBG) parameters. Kim et al 6 compared experimental GMAW weld bead geometry results with those obtained from heat-transfer and regression models. Kim et al 7 applied modified Taguchi method to determine process parameters for optimum weld pool geometry in TIG welding of stainless steel. Tarng et al 8,9 determined optimum process parameters for submerged arc welding (SAW) in hard facing using grey-based Taguchi method. New trend in manufacturing processes parameters optimization is to use evolutionary algorithms such as genetic algorithm (GA) 10 and simulated annealing(SA) 11 . Other search methods have also been used for this purpose 12 . Along this line, SA Algorithm is a well known evolutionary method successfully adopted in different areas 13 . This study proposed a SA approach to establish relationships between process parameters (inputs) and responses (outputs) in metal active gas (MAG) welding using RA carried out on data collected as per Taguchi design of experiments (DOE). Experimental For modeling and optimization of MAG welding process (Fig. 1), a consumable electrode is used as filler with an active gas shielding to protect molten metal from oxidation. Important input parameters in MAG are welding speed (S), welding voltage (V), wire feed rate (F), nozzle-to-plate distance (D) and torch angle (A), whereas output parameters (responses) are bead height (BH), bead width (BW) and bead penetration (BP). Three levels were considered for input process parameters (Table 1) and 54 combinations of input process parameters were considered for Taguchi DOE