Indian Journal of Engineering & Materials Science Vol. 21, April 2014, pp. 149-154 Development of regression models and optimization of FCAW process parameter of 2205 duplex stainless steel G Bansal Rajkumar a * & N Murugan b a Department of Mechanical Engineering, United Institute of Technology, Coimbatore 641 020, India b Department of Mechanical Engineering, Coimbatore Institute of Technology, Coimbatore 641 035, India Received 4 June 2012, accepted 3 December 2013 Welding input process parameters are playing significant role to determine the weld bead quality. This paper presents an experimental design approach to optimizing the input process parameter of flux cored arc welding (FCAW) of 2205 duplex stainless steel butt weld. FCAW input process parameters such as welding current, welding speed and open circuit voltage are taken as an important parameter to determine the quality of weld. The experiment is conducted based on three factors and five levels of central composite rotatable design with full replication technique. Regression models are developed to determine optimal parameters. Weld metal area of the weld bead, an important bead parameter is optimized (minimized) by keeping weld bead width and reinforcement as constraints to obtain quality weld. The optimized parameter is determined by Excel solver and validated by conducting experiments. The results reveal the interdependence of weld bead quality parameters in controlling weld metal area of the bead to improve the weld quality. Keywords: FCAW, Optimization, Weld metal area, Response surface methodology, Regression models Various methods of welding processes like shielded metal arc welding (SMAW), gas tungsten arc welding (GTAW), gas metal arc welding (GMAW), and flux cored arc welding (FCAW) are used for joining of stainless steels. FCAW has been extensively employed in industries 1 because it gives better welds with consistent mechanical and metallurgical properties, few weld defects and high deposition rate can be operated with marginal skills. It is suitable for mild and low alloy steel, some high nickel alloy and stainless steels 2. Stainless steels are popularly used in industries because of its corrosion resistance. Types of steels can be identified based on the microstructure and major crystal phase. Duplex stainless steel (DSS) is the mixture of austenitic and ferritic stainless steels. Hence, it has inherent properties such as high fatigue strength, good pitting resistance and weldability, high mechanical strength and good in corrosion resistance due to high content of chromium 3. 2205 DSS is widely employed in desalination plants, pulp industry, bridges, pressure vessels cargo tanks, pipe systems in chemical tankers, and heat exchangers due to both excellent corrosion resistance and high strength. *Corresponsing author (E-mail: bansalgnanasundaram@gmail.com) FCAW input process parameters are welding speed, arc current, arc voltage, open circuit voltage (OCV), nozzle to plate distance (NPD), electrode protrusion, gas flow rate and preheat temperature 4. Among these, selections of significant parameters are very important to obtain the desired bead geometry and good weld quality. Input process parameters such as current, welding speed, open circuit voltage and the effects of bead geometry are discussed by conducting experiments based on three factors and five level central composite rotatable design matrix with full replication technique. FCAW has been employed to weld 2205 duplex stainless steel. Regression models have been developed and the effects of process parameters on the weld bead geometry are estimated and presented in graphical form. The developed regression models will be useful in selecting the appropriate FCAW process parameters to achieve the desired weld quality. Generally, suitable welding parameters can be selected by the skilled welders and engineers based on their experience and by trial and error method. After conducting trial runs, they inspect whether it meets the joint requirement or not. This type of process is more expensive and time consuming. The above constraints have been overcome by the design of experiments, which will help to develop the
150 INDIAN J ENG MATER SCI., APRIL 2014 correlation between input process variables and output bead quality. Design of experiments (DOE) technique also helps to optimize the welding process to achieve desired bead quality 4. Urena 5 developed the optimum welding condition for joining 2205 DSS using plasma-arc welding. Ku et al. 6 welded 2205 duplex stainless steel by electron beam welding and analyzed the mechanical properties, microstructure and corrosion properties of the weldment. Kannan and Murugan 7 observed the effects of FCAW process parameters on DSS clad quality. Laser beam welding parameters were optimized by Reisgen et al. 8 and they found that, RSM can be considered as a powerful tool in experimental welding optimization. As the weld bead quality depends on the process parameters, it is essential to study the effects of process parameters on weld quality. Experimental Procedure In the present work, 2205 DSS plates were joined with E2209T1 filler wire by FCAW process. The materials were obtained from Outokumpu Stainless AB, Sweden. The chemical compositions of both base and filler are given in Table 1. A flux cored arc-welding set-up is shown in Fig. 1. It consists of a power source, filler wire feeding unit with shielding gas flow control and welding gun. A semi-automatic welding manipulator helps to deposit the filler metal on the required area by controlling two axes. FCAW parameters of welding speed, current and open circuit voltage were considered to conduct trial runs by varying one process parameter and keeping the other two parameters as constant 9. The working range was decided based on the appearance of weld bead based on smooth continuous bead, absence of porosity, undercut during visual inspection. The upper limit was coded as 1.682 and the lower limit was coded as - 1.682, the intermediate values can be predicted from the relationship of X i = 1.682 [2X - (X max + X min )] / (X max - X min ) (1) Where X i is the expected coded value of the variable X, X is any variable between the value of X min to X max 10. Selected levels of the FCAW control process parameters are given in Table 2. 2205 DSS plates of size 150 50 6 mm 3 were butt welded with a root gap of 2 mm using flux cored DSS (E 2209T1-4/1) welding wire of 1.2 mm diameter. Shielding gas containing 75% Argon plus 25% CO 2 with a gas flow rate of 20 L/min was maintained 11. Plates were welded on both sides by keeping electrode-to-work at an angle of 90. Inter pass welds temperature of 150ºC was maintained and measured with the help of an infrared non-contact digital thermometer. The welding trials were conducted as per the design matrix at random to avoid the systematic errors creeping into system 12. The Box- Wilson Central composite design matrix is commonly used for constructing a second order polynomial regression model. The selected design matrix is shown in the Table 3. It consists of 20 runs. Fig. 1 FCAW experimental set-up Material Table 1 Chemical compositions of base metal and filler wire Chemical composition, % by weight C Si Mn P S Cr Mo Ni N2 Cu N 2205 (Base metal) 0.02 - - - - 22 3.1 5.7 - - 0.17 E2209T1-4/1 0.023 0.76 1.03 0.024 0.002 23.14 3.05 9.22 0.13 0.09 - Table 2 Control parameters and their levels Sl. no. Parameters/units Factor Levels -1.682-1 0 1 1.682 1 Welding current (A) 170 182 200 218 230 2 Welding speed (cm/min) 25 27 31 35 37 3 Open circuit voltage (V) 28 30 32 34 36
RAJKUMAR & MURUGAN: FLUX CORED ARC WELDING OF 2205 DUPLEX STAINLESS STEEL 151 All specimens were prepared by a standard metallurgical procedure to measure the bead dimensions. The weld bead profile was revealed by etching the specimens with ferric chloride and bead parameters were traced with the help of a profile projector. CAD software was used to measure bead width, reinforcement and weld metal area. Typical cross-section of a weld bead is shown in Fig. 2. The observed values of bead geometry are given in Table 3. Development of Regression Model The response surface function representing parameters can be expressed as 13 : For three factors, the above polynomial is expressed as Y = a 0 +a 1 I +a 2 S+a 3 V+a 12 IS + a 13 IV+ a 23 SV+ a 11 I²+ a 22 S²+ a 33 V² (4) where a 0 is the free term of the regression equation, the coefficients a 1, a 2 and a 3 are the linear terms, the coefficients a 12, a 13 and a 23 are the interaction terms and the coefficients a 11, a 22 and a 33 are the quadratic terms. The coefficients were determined by Quality Y = f (X 1, X 2, X 3 ) (2) Where Y is the response such as width, reinforcement, weld metal area, X 1 is the welding current (I), X 2 is the welding speed (S) and X 3 is the open circuit voltage (V). The second-order polynomial representing the response for K factors is given as 13 : k k k Y = a + a X + a X X + a X i= 1 = 1 i=i j 0 i i ij i j ii i i.j (3) 2 Fig. 2 A typical cros-section of a welded plate showing weld bead profile Specimen no. Table 3 Design matrix with observed values of bead dimensions Design matrix Weld bead parameters I (A) S (cm/min) V (V) W (mm) R (mm) WA (mm 2 ) 1-1 -1-1 12.74 2.28 707.23 2 1-1 -1 12.27 2.99 855.4 3-1 1-1 11.23 1.46 578.14 4 1 1-1 13.17 2.18 772.56 5-1 -1 1 13.48 2.19 766.99 6 1-1 1 15.63 2.44 1015.64 7-1 1 1 13.33 2.22 788.49 8 1 1 1 12.87 1.86 721.23 9-1.682 0 0 12.59 1.65 680.2 10 1.682 0 0 15.46 2.47 953.22 11 0-1.682 0 14.95 2.35 904.69 12 0 1.682 0 13.21 1.59 716.4 13 0 0-1.682 11.55 2.3 697.78 14 0 0 1.682 14.01 1.84 828.69 15 0 0 0 14.12 1.92 826.39 16 0 0 0 15.38 2.08 873.23 17 0 0 0 13.88 1.61 758.96 18 0 0 0 13.28 2.24 742.33 19 0 0 0 14.37 2.19 753.08 20 0 0 0 12.48 2.16 716.01 I-current, S-speed, V-Open circuit voltage, W-Width, R-Reinforcement, WA- Total area
152 INDIAN J ENG MATER SCI., APRIL 2014 America (QA) software. Regression models were developed using the coefficients. The developed models in coded form are given below. Bead width, W = 13.933 + 0.585 I - 0.472 S + 0.734 V - 0.071 I 2-0.051 S 2-0.513 V 2-0.025 IS + 0.026 IV - 0.289 SV (5) Reinforcement, R = 2.028 + 0.198 I - 0.253 S - 0.072 V + 0.045 I 2 + 0.014 S 2 + 0.048 V 2-0.073 IS - 0.192 IV + 0.135 SV (6) Weld metal area, WA = 787.387 + 71.986 I - 58.686 S +43.872 V + 6.028 I 2 + 3.849 S 2-12.873 V 2-33.706 IS - 20.149 IV - 7.622 SV (7) Developed models have been verified by drawing the scatter diagrams plotted between measured and predicted values of W, R and WA, few of them are shown in Fig. 3. The adequacy of the developed models was tested using the analysis of variance (ANOVA) technique that is given in Table 4. F ratios are greater than the tabulated values at the 95% confidence level and hence the models are adequate. Fig. 3 Scatter diagram of (a) reinforcement model and (b) weld metal area model Conformation test reports to check the accuracy of the developed models After the development of regression models, their accuracy was determined by conducting conformation test runs using the same experimental setup. The conformation test runs were conducted with different sets of input process variables other than that presented in the design matrix. The responses were measured and compared with predicted values of bead geometry and it is given in Table 5. The results show that models are acceptable. Optimization of the FCAW process parameters In welding, the weld metal area is an important parameter to be considered for optimization. By understanding the interdependence of various weld bead parameters, weld metal area is controlled by the other parameters. Hence the weld metal area, if optimized (minimized), obviously minimizes other bead quality parameters such as width and reinforcement. Minimizing the size of the weld bead obviously gives the following benefits. It reduces the welding cost through reduced consumption of consumables such as electrodes and flux. The heat input and energy consumption are reduced. The welding productivity could be improved through a high welding speed. Because of these advantages, the weld metal area should be optimized, having other bead parameters as constraints, rather than optimizing all the bead parameters individually. Khan et al. 14 used experimental design approach to optimize laser Parameters Table 4 ANOVA for testing adequacy of the models Width Bead geometry Reinforcement Weld metal area Regression SS 18.78 1.92 153193.5 DF 3 3 3 Residual SS 10.382 0.583 55315.04 DF 16 16 16 Lack of fit SS 5.515 0.307 26198.21 DF 10 9 10 Error term SS 4.867 0.276 29116.84 DF 5 5 5 F-ratio (calculated) 5.727 17.924 4.057 F-ratio (tabulated) 3.24 3.24 3.24 R 2 value 0.644 0.767 0.735 Adjusted R 2 value 0.549 0.684 0.664 Remarks Adequate Adequate Adequate SS: Sum of Squares, DF: Degree of Freedom, SS/DF = Mean square, F-ratio = Mean sum of squares for regression / Mean sum of squares of error
RAJKUMAR & MURUGAN: FLUX CORED ARC WELDING OF 2205 DUPLEX STAINLESS STEEL 153 Table 5 Conformation test results to check the accuracy of the developed models Input / response parameters Input parameters Conformation test no. 1 2 3 Current (A) 0 1 0 Speed (cm/min) 1 0 1 OCV (V) -1 1 0 Measured values of bead geometry Width (mm) 12.98 13.87 13.08 Reinforcement (mm) 1.95 1.88 1.61 Weld metal area (mm 2 ) 665.09 893.02 698.5 Predicted values of bead geometry Width (mm) 13.02 13.53 13.41 Reinforcement (mm) 1.97 1.96 1.59 Weld metal area (mm 2 ) 683.43 876.25 712.55 % Error Width -0.31 2.54-2.46 Reinforcement -1.22-3.84 1.32 Weld metal area -2.68 1.91-1.97 Average -1.403 0.203-1.036 % Error = [(Measured value - Predicted value) / Predicted value] 100 beam welding process parameter with the help of regression models. In order to carry out the optimization, regression equations are required for predicting the values of weld bead geometry such as bead width, reinforcement. The developed models in the regression analysis were used to optimize the FCAW process to achieve minimum weld metal area. Optimization of the FCAW Process Parameters Using Excel Solver Gunaraj and Murugan 15 used the regression model as the objective function for optimization. The objective function selected for optimization was the weld metal area. The other bead parameters like bead width, reinforcement were given as the constraints of the objective equation and the input process parameters were not to exceed their minimum and maximum range. Microsoft Excel 2007 solver was used to carry out the optimization. Solver works with a group of cell, which directly or indirectly related, to the formula (Objective equation from regression modelling) in a target cell. Solver manages the values in the changing cells called the adjustable cells to produce the result. Constraints are used to restrict the values of the variables used in the objective function. The objective function for minimization The objective function and the constraints are given as: Weld metal area, ƒ (χ) = 787.387 + 71.986 I - 58.686 S +43.872 V + 6.028I 2 + 3.849 S 2-12.873V 2-33.706 IS - 20.149 IV - 7.622 SV (8) The given objective function was optimized subject to the following constraints: -1.682 < Welding current < 1.682-1.682 < Welding speed < 1.682-1.682 < Open circuit voltage < 1.682 The constraint equations are Bead width, W = 13.933 + 0.585 I - 0.472 S + 0.734 V - 0.071 I 2-0.051 S 2-0.513 V 2-0.025 IS + 0.026 IV - 0.289 SV 10 (9) Reinforcement, R = 2.028 + 0.198 I - 0.253 S - 0.072 V + 0.045 I 2 + 0.014 S 2 + 0.048 V 2-0.073 IS - 0.192 IV + 0.135 SV - 2.1 (10) The set of optimal solutions for minimizing weld metal area obtained from the Excel solver are as follows Set 1 I = - 1.682, S = - 1.406, OCV = - 1.682 W = 10.4675 mm, R = 1.87 mm, WA = 617.813 mm 2 Set 2 I = - 1.682, S = 1.682, OCV = - 1.682 W = 9.6725 mm, R = 1.307 mm, WA = 586.643 mm 2 Results and Discussion Conformation test with optimized process parameters The accuracy of the optimized FCAW process parameters was determined by conducting conformation test runs using the same experimental setup. Weld trials were conducted for the optimal parameter settings. The cross-section of the optimized weld bead is shown in Fig. 4. A comparison was made between the predicted and actual values of bead parameters and the results of the conformation test with optimised process parameters are given in Table 6. From this table it is found that the average error is less than 1 %.
154 INDIAN J ENG MATER SCI., APRIL 2014 conformation test. Optimized parameters are obtained and verified by conducting optimization run. The obtained optimized input parameters help to get optimum weld bead with the determined quality. Fig. 4 Cross-section of a optimized specimen (a) Set 1 and (b) Set 2 Table 6 Results of the conformation test with optimized process parameters Trail no. Optimized FCAW process parameters Weld metal area (mm 2 ) % error I S OCV Measured Predicted 1-1.682-1.407-1.682 603.853 593.271 1.784 2-1.682 1.682-1.682 586.643 587.115-0.080 Average error 0.852 Conclusions Based on the flux cored arc welding process parameters taken in this study the following conclusions are drawn. Regression models for weld bead geometry such as width (W), reinforcement (R), weld metal area (WA) have been developed and their adequacy has been confirmed by ANOVA. The accuracy of the models is determined by conducting a References 1 Juffus L, Welding Principles and Applications, (Delmar publisher, New York), 1999. 2 Flux Cored Welding, Welding Procedures & Techniques, (American Metallurgical Consultants), 2006. 3 KimY S & Eagar T W, Weld J, 72(7) (1993) 279-287. 4 Benyounis K Y & Olabi A G, Adv Eng Software, 39 (2008) 483-496. 5 Urena A, J Mater Process Technon, 182 (2007) 624-631. 6 Ku J S, Ho N J & Tjong S C, J Mater Process Technon, 63 (1997) 770-775. 7 Kannan T & Murugan N, J Mater Process Technon, 176 (2006) 230-239. 8 Reisgen U, Schleser M, Mokrov O & Ahmed E, Opt Laser Technol, 44 (2012) 255-262. 9 Gupta V K & Parmar R S, J Inst Eng (India), 70 (1989) 67. 10 Palini P K & Murugan N, J Mater Process Technon, 172 (2006) 1-10. 11 How to weld 2205 duplex stainless steel, Outokumpu Stainless AB, Avesta Research Centre, Sweden. 12 Murugan N & Parmar R S, Weld J, 10 (1997) 391-402. 13 Kathleem M Carley & Natalia Y Kanneva, Response Surface Methodology, CASOS Technical Report, October 2004, CMU-ISRI- 04-136 14 Khan M M A, Romoli L, Fiaschi M, Dini G & Sarri F, Opt Laser Technol, 43 (2011) 158-172. 15 Gunaraj V & Murugan N, Weld J, 79 (2000) 331-338.