Optimization of TIG welding process parameters on 304 austenitic stainless steel sheet metal using fuzzy logic based Taguchi method

TIG welding can be used to produce excellent weld quality and precise welding operation for sheet metals. The aim of this study was to get the best welding parameters for enhancing ultimate tensile strength, bending strength and Rockwell hardness of the butt-weld joint. The experimental work was used, and the experiment was carried out on 2 mm thickness of 304 austenitic-stainless steel sheet metal using the L9 orthogonal array of the Taguchi design. Automated TIG welding fixture was developed to control the welding speed accurately. The selected welding parameters were welding speed, current, voltage and gas flow rate with their three levels. Based on the fuzzy logic based Taguchi method, the best optimal levels of parameters were found at the values of 110 A of current, 13.86 cm/min of speed, 17.5 V of voltage and 7.5 L/min of gas flow rate. The analysis results of ANOVA showed that gas flow rate and current were found as the significant factors, and the contributions of the gas flow rate, current, speed and error were 47.63%, 34.34%, 16.49% and 1.54%, respectively. According to the confirmation tests, the multi response performance index mean value of the confirmatory test of 0.6068 was found between the 95% confidence interval of 0.5028 and 0.7314, and the maximum ultimate tensile strength, bending strength and Rockwell hardness were obtained 614.8 MPa, 765.32 MPa and 95.3 HRB, respectively.


Introduction
Tungsten inert gas (TIG) welding is an essential arc welding technology that uses an inert gas for welding pool arc shielding and a tungsten non-consumable electrode to produce extremely excellent weld quality and precise welding operation.An arc is produced between a non-consumable tungsten electrode and a base metal.When a groove has to be filled, a thin wire of filler metal is melted directly in the molten weld pool.Because of its excellent electrical and high melting temperature conductivity, tungsten is the ideal non-consumable electrode material.A small proportion of lanthanum, cerium, thorium, or zirconium is added to tungsten electrodes to increase current carrying capacity, arc striking ability, and stability [1,2].TIG welding is the popular welding that is widely employed in modern industries that need high-quality welding, such as aviation, vehicles, steam, chemical, nuclear reactors, food, shipbuilding, and bridge construction to join similar or dissimilar materials [3][4][5][6].
Welding sheet metal is one of todays most difficult engineering tasks in determining the best welding parameters' combinations to attain the required weld quality [7,8].In order to achieve the welded product's reliability, productivity, and quality, it is necessary to optimize the welding parameters, such as voltage, speed, gas flow rate and current [9].
Researches had been conducted on the welding of stainless steel metals.Some of the studies were stated as the followings.Influences of welding parameters on stainless steel 304 sheet metal that were welded using the TIG welding process were studied to improve the penetration depth, hardness and ultimate tensile strength [10].As per this study, the optimal ultimate tensile strength of 556 MPa and Vickers hardness of 100 HV were found using the Taguchi method at current of 150 A, gas flow rate of 12 L/hr, welding speed of 3.2 mm s −1 , and filler rod diameter of 2.4 mm.Current, welding speed and electrode diameter were the major welding factors that influenced the quality of the weld joint.Effects of the pulse TIG welding parameters on 3 mm thickness of stainless steel 304 sheet metal were also studied by considering welding current, welding speed and gas flow rate as input parameters to improve weld quality [11].The ultimate tensile strength of the butt weld joint was improved for those parameters using the Taguchi method, and its result was to be 448 MPa at gas flow rate of 9 L/min, current of 90 A, and travel speed of 5.2 mm s −1 .The most significant factors were also found to be welding current and travel speed.The 3 mm thickness of austenitic stainless steel AISI 304L was welded using the TIG welding process, and its weld bead micro-hardness and ultimate tensile strength of the joint were studied by considering welding current, welding speed and gas flow rate as control factors using the Taguchi method based on fuzzy logic techniques [12].The best maximum weld bead micro-hardness and ultimate tensile strength were achieved with 135 A of welding current, 190 mm/min of welding speed, and 4 L/min of gas flow rate.The welding speed, gas flow rate and welding current had highest, medium and lowest impacts on the weld quality, respectively.Another study [13] was conducted on 3 mm thickness of stainless steel 304 that was welded using the TIG welding process.In this study, argon was used as shielding gas, and gas flow rate, voltage and welding current were also used as control parameters.At the current of 130 A, voltage of 20 V and gas flow rate of 25 L/ min, the greater tensile strength and hardness of 631 MPa and 211 HV were achieved, respectively.Significant parameters were determined using ANOVA; as a result, welding current was the most significant factor for tensile strength, whereas the gas flow rate was the most significant factor for hardness.Besides, the effects of the TIG welding control factors of welding current, gas flow rate, root face thicknesses and welding speed on depthof-penetration (DOP) of welded stainless steel 304 plate having 6 mm thickness was studied [14].This experimental study was carried out using the Taguchi design trials technique.110 A of welding current, 10 L/ min of gas flow rate, 1.5 mm of root face, and 31.578mm/min of welding speed were found to be the best setting of levels for the greatest depth of penetration (DOP), with a response value of 5.6961mm.The contributions of welding current, gas flow rate, root face and welding speed to the DOP were 46%, 32%, 6.94% and 14.7%, respectively.
Researchers investigated effects of the TIG welding process parameters, such as welding current, welding speed, arc voltage, shielding gas flow rate, filler rod diameter and root face thickness.Aside from that, product characteristics, like bead shape and weld mechanical properties had been studied.the TIG welding process can be used for welding thin to thick metals; however, the majority of researches were confined to thick metal plates rather than thin sheet metals of 304 stainless steel.Furthermore, most optimizations of control factors were performed on single response rather than combinations of multiple responses.
The aim of this study was to optimize the TIG welding process parameters on 2 mm thickness of 304 austenitic stainless steel sheet metal since there was not any carried out research on such thickness of sheet metal to optimize the TIG welding parameters for multiple responses simultaneously.The 304 austenitic stainless steel sheet is the most commonly utilized stainless steel grade in the world, and it is used almost everywhere sectors and industries, such as food, chemicals, textiles, aerospace and transportation since it is mildly and sturdy corrosion resistant.As a result, this study addressed the observed research gaps in order to improve this sheet metal welding qualities by optimizing the TIG welding parameters using the fuzzy logic based Taguchi method.

Materials
Stainless steel (SS) comes in a variety of grades, including austenitic, ferrite, martensitic, and duplex.Among them, austenitic stainless steel 304 of 2 mm thickness was selected.The selection criterion of 304 austenitic SS having 2 mm thickness was based on the literature reviews, since this sheet metal was not examined before with a combination of as in this study's welding parameters and responses.The target metal was also chosen because it is commonly employed in the manufacture of industrial and transportation equipment.The chemical compositions of this base metal were investigated using metal analytic spectrometer machine having Model-MAXx LMD06; Make-Spectro Analytical; Germany, and obtained results are shown in table 1.
An automated welding fixture was designed as shown its CAD model on figure 1(a).This fixture has components of screw shaft, stepper motor (Bipolar, 200 Steps per Revolution with, 5.1V, 1.1 A/Phase), supporters, welding torch holder, moving tractor and ball bearing.The frames of the fixture were made from mild steel and angle iron.Its overall dimensions were reduced to 500 × 300 × 250 mm by grinding, and the joints were welded using a manual metal arc-welding equipment.In this study, the welding speed was controlled automatically using this developed fixture as shown its control units on figure 1(b).The control units were managed using Arduino IDE 1.8.19 coding program and stepper motor.The length of the horizontal movement frame was specified for liner movement.The horizontal stand component was connected to the top of the vertical plate, and the vertical stand was also bolted to the base plate.The horizontal torch movement was controlled by Arduino program using a nut and screw mechanism.The experimental configuration was designed to manage the linear travel speed of the torch along the weld pad center line to move forward and backward.After fabrication and assembling of the automated welding fixture as shown its experimental setup on figure 1(c) with welding torch and speed control units, series of trial tests were carried out before conducting the  actual experiment in order to determine the acceptable parameter range for welding of the 304 austenitic stainless steel sheet metal.

Parameters and orthogonal array sections
In this study, the welding parameters were determined according to literature review and the TIG welding of stainless steel surveys.The most important parameters for the TIG welding equipment were selected using the Pareto diagram as shown in figure 2. The first four parameters were chosen as control welding parameters since they cumulatively contribute more than 80%, each having its unique effect on the weld-joint quality, while the remaining parameters were kept as constants.Due to these, the control factors were current, speed, voltage and gas flow rate for welding the selected metal and forecasting ultimate tensile strength, bending strength and Rockwell hardness strength.The levels of selected welding parameters shown on table 2 were also chosen based on the specifications of the TIG welding equipment having Model-EURO TIG 200 AC/DC; Make-Helvi; Italy, utilized in this study and by reviewing previous researches that were in the same context as the welding parameter decision [2,15,16].The four control parameters had three levels; besides, there were also the fixed TIG welding parameters that are listed on table 3 during this experimental work.Four welding parameters with their three levels were considered for this study, with two alternatives L9 and L27 orthogonal arrays were being viable under Taguchi design.The L27 might have higher prediction accuracy; however, L9 was provided to minimize the number of runs and cost [2,[15][16][17]; as a result, the selected orthogonal array was L9 that is shown on table 4.

Testing of the welded sheet metal specimens
Among the mechanical properties, ultimate tensile strength, bending strength and Rockwell hardness of the weldment of 304 austenitic stainless steel were selected as responses in this study by considering importance to the industries, analysis method, joint design types, applicability, and availability of experiment [2].
Tensile testing is one of the most effective criterion for determining weld quality in any welding experiment [18].As a result, the weldment's tensile strength was determined using a tensile test.A computer-controlled hydraulic universal testing equipment having Model-WAW-600D; Make-Sunpoc; China, was used to perform the tensile tests.Tensile test specimens were built to ASTM requirements after welding [19][20][21].The samples were machined and used for the testing purpose as shown on figure 3.
An external load was applied to the welded connection region on the flat surfaces in a bending test, causing the specimen to bend into a U-shape using a computer-controlled hydraulic universal testing machine having Model-WAW-600D; Make-Sunpoc; China, with three-points bend test.As a reason, evaluating stability for mechanically-joined sheet metal components demanded a detailed examination of sheet metal bendability [22].The bending tests were performed on the weld to determine the existence of a detrimental flaw in the weld metal or whether it had sufficient ductility that matched the base metal [23].When defects in the material arise while being subjected to strong stresses from the bend test, the material might rip locally, resulting in specimen failure.Bend test standard welded butt connections were used as standard in bend test of metallic materials [2,21].The prepared and tested specimens for bending strength are shown on figure 4.
Welding-induced metallurgical changes, in addition to brittleness and relative crack sensitivity to structural forces, are shown by hardness testing.The Rockwell hardness test was used to compare the depth of penetration produced by an indenter under a heavy load against the penetration depth made by a preload to evaluate hardness [24].The Rockwell hardness test was performed using machine having Model-MAT10RAB, Make-Material testing machines services; India.And, it was done based on the minor and major load principle, in which a small load of 10kg was applied to the sample to minimize surface preparation and minor defects, and then a high weight of 100 kg was applied to the sample according to scale B for a duration of 15 seconds.The prepared and tested specimens for Rockwell hardness are shown on figure 5.

Optimization by fuzzy logic based Taguchi method
According to the reviewed studies, the Taguchi method was mostly used to optimize welding parameters on stainless steels.The Taguchi technique is a potent problem-solving tool that may be used to improve a process performance without requiring a huge number of experiments.However, the majority of Taguchi applications that had been reported so far have been focused on improving a single performance metric [2,16].The Taguchi method can be coupled with grey relational analysis, fuzzy logic and others to optimize multiple responses [2,12,16,25].However, researchers would require to find additional method for determining weighting factors responses for grey relational analysis (GRA), and principle component analysis (PCA) is mostly used along with GRA.PCA has also limitation on determining weighting factors for two responses, since it mostly gives 0.5 value for each of two responses that would be obtained from eigenvalues and eigenvectors.
The fuzzy logic is a form of expert system for determining the optimal values from a set of input and output variables [26].Generally, fuzzy controller consists of four modules [12].
• Fuzzification-fuzzifier fuzzifies the signal to noise ratios using membership functions created by the response.
To describe these fuzzy sets, fuzzy membership values based on various membership functions are utilized.
• Fuzzy rules base expert modeling expertise and experience, as well as relating fuzzy model input variables to output variables, are both enhanced by fuzzy rules.• Fuzzy inference by executing fuzzy reasoning on fuzzy rules, the inference engine generates a fuzzy value.In a fuzzy rule, it is utilized to determine the rule output from the given rule input information.
• Defuzzification creates a multi-response performance index from a fuzzy value.
The fuzzy logic tool is an outstanding predictor, and its generated model had shown to be highly useful in handling works, energy, money and saving time that would otherwise had been wasted in pre-welding   operations [27].The Taguchi and the fuzzy logic were used to determine the best levels of welding parameters of the TIG welding process to optimize output responses to the required quality of weld in this study.To merge multiple responses into a single response during optimization, the fuzzy logic coupled with the Taguchi method was used [28].In such approach, the problem of optimizing numerous parameters can be solved using coupling method of the Taguchi and fuzzy logic.Figure 6 depicts the schematic representation of the optimization flowchart of the fuzzy logic based Taguchi method that was implemented in this study.

Results and discussions
Trial experiments were conducted using L9 orthogonal array of the Taguchi design, and values of the responses of ultimate tensile strength, bending strength and Rockwell hardness were measured and their results are shown on table 4.

Visual inspection
The types of faults that were developed at the weld line were also observed using visual assessment of surface failure.Weld defects can be divided into several kinds based on their size, position, and form in a specific environment.The visual inspection was carried out to assess the efficacy of the TIG welding on the butt weld joints, and defects and their causes were interpreted as shown on figure 7.As show on figure 7(a), crack was occurred due to air bubbles and insufficient heat at the starting of left side.Surface porosity was also observed on experimental trial as shown on figure 7(b), and its possible causes were air bubbles and gas trapped along the weld zone.Experimental trials 3, 4 and 6 which are shown on figures 7(c), (d) and (f), respectively, did not have visually observed defects.However, spatter defect was observed on figure 7(e),  and its possible causes were spilling of the molten metallic drops during welding that stuck to the welding surfaces.Besides, over lap defect was observed due to excessive welding current and slow welding speed on experimental trial 7 as shown on figure 7(g), and under filling was occurred at the end of the right side of the image on experimental trial 8 which is shown on figure 7(h).Necking defect was occurred on experimental trial 9 which is shown on figure 7(i), and its possible cause was insufficient heat due to high speed at starting point.
Heat input has significant effects to achieve sound microstructure and weld quality, and it is the function of current, welding speed, voltage, gas flow rate, arc length and heat transfer efficiency.The cooling rate from the peak temperature obtained during the welding cycle determines the final metallurgical structure of a weld zone (heat-affected zone and fusion zone).For diffusion controlled transformations, the rate of cooling influences the coarseness or fineness (grain size) of the resulting solidification structure and homogeneity, as well as the distribution and form of the phases and constituents in the microstructure of both the heat-affected zone and fusion zone.If cooling rates are excessively high in some steels, hard, untempered martensite can form, embrittling the weld and increasing the vulnerability to hydrogen embrittlement [29].In this study, the arc length was constant and the welding environment was the same.And, it was assumed that as gas flow rate  increased, cooling rate would also increase.In contexts of these, locations and types of fractures that were observed on tensile tested specimens on figure 3(b) are briefly summarized in table 5 with their possible causes [15].

Effects of welding parameters on responses 3.2.1. Effects of current
As current increased, ultimate tensile strength increased.So, the increment of current showed that good outcome for this response.However, bending strength and Rockwell hardness did not show linear relationship with current.When current increased, bending strength decreased and then increased, whereas Rockwell hardness increased and then decreased.These effects of current for the three responses are shown on figure 8.

Effects of speed
As shown on figure 9, bending strength, ultimate tensile strength and Rockwell hardness did not show linear relationship with speed since when the speed increased, bending strength increased and then decreased but both ultimate tensile strength and Rockwell hardness decreased and then increased.

Effects of voltage
As shown on figure 10, bending strength, ultimate tensile strength and Rockwell hardness did not show linear relationship with voltage.When welding voltage increased, both ultimate tensile strength and bending strength decreased and then increased, whereas Rockwell hardness increased and then decreased.

Effects of gas flow rate
As shown on figure 11, all ultimate tensile strength, bending strength and Rockwell hardness did not show linear relationship with gas flow rate.As gas flow rate increased, ultimate tensile strength, bending strength and Rockwell hardness increased and then decreased.All three responses were observed that from figure 8, figure 9, figure 10 and figure 11, they did not have the same effects as the values of levels of welding parameters increased.As results, optimization of the TIG welding parameters using multi-objective method was mandatory to address the effects.Therefore, the fuzzy logic based Taguchi method was applied to optimize the TIG welding parameters and its implementation phases were summarized on figure 6.

Signal to noise ratios
The Taguchi method determines the best production design choice based on the ratio of signal to noise (S/N) ratio [30].The response must be on the higher side for a high-quality weld.As a result, the higher S/N ratio indicates the lower variance in response, which is desirable [31].Among available S/N ratios of lower is better, larger is better and nominal is best, larger is better (η H ) was taken in this research since the responses were needed to have maximum values, and its equation (1) [32,33] was used to analyze experimental results which are shown on table 4, and the analyzed S/N ratios are also listed on table 6.
Where, V ij is the response variable value of the j th performance characteristic in the i th experiment and R is number of experimental replications.

Membership functions of input and output variables
The Mamdani type fuzzy interference system (FIS) was used to generate a membership function with three inputs and one output.The multi response performance index (MRPI) of the optimal combination of ultimate tensile strength (UTS), bending strength (BS) and Rockwell hardness (HR), as well as their language terms and membership functions, were determined as per the flowchart shown on figure 12. Twenty-seven rules were used to determine the MRPI of the best combination of the three responses, as well as their linguistic terms and membership functions.
In the rule viewer, the output parameters were examined by altering the values of the three inputs from low to high.The fuzzy logic MATLAB tool box, which determined the input and output variables, is shown in figure 13, and it was the fuzzy model of TIG welding process.The fuzzy logic was based on calculating the fuzzyset, which reflected the potential values of the variables.For all linguistic phrases in all variables, the triangle membership function with values ranging from 0 to 1 [34,35] was utilized.Low, medium, and high membership functions were chosen for each input and output variable.Figures 14, 15 and 16 show the fuzzy logic MATLAB tool box, which determined the signal to noise ratios of inputs of UTS, BS and HR with three membership function variables, respectively.
The anticipated output parameter was separated into five levels, as illustrated in figure 17: L = low, LM = low-medium, M = medium, MH = medium-high and H = high.Here, the inference rules and decision bases of knowledge in the MATLAB software were programmed.The membership function for input variables such as UTS, BS, and HR, as well as the output MRPI, had been incorporated to a new fuzzy interference system.Twenty-seven 'IF AND THEN' rules were constructed and added between three inputs and one output membership function as indicated in figure 18.The fuzzy criteria were founded on the principle of 'the bigger, the better.'The input and output responses structure in the ANFIS tool were modeled using MATLAB in this study as shown in figure 19.The process variables were entered first, followed by the triangular membership function for input parameters and the linear membership function for output responses in the command window.Figure 19 depicts the established ANFIS structure, which had three membership functions for all input parameters and five membership functions for the output response.The three input parameters were displayed in the first three columns, while the output parameter was displayed in the final column in the rule viewer as shown in figure 20, and results of MRPI and Y model are shown on table 6.
Multiple linear regression modeling were developed for comparing the curve fit data with fuzzy logic modeling.As demonstrated in figure 21, the models of fuzzy and multi linear regression comparison graph showed that more correlated to each other with best curve fit lines.The comparison of errors between the fuzzy logic and multi linear regression model showed that all were below 10% which was in the acceptable ranges.

Optimal levels of TIG welding parameters
As indicated in table 7, the average of the response data was calculated using Minitab software at three different levels for varied values of current, speed, voltage, and gas flow rate [2,16,25], and their primary effects of plots for MRPI are given in figure 22. the greatest MRPI value for each column of factors denoted the best level for that factor.As a result, the best optimal levels of the TIG welding parameters in the supplied model that generated the MRPI values which are shown on table 7 and figure 22 was A 3 B 1 C 2 D 2 , and it implied that the optimal levels of TIG welding parameters were obtained current at 110 A, speed at 13.86cm/min, voltage at 17.5 V, and gas flow rate at 7.5 L/min.The TIG welding process was influenced more by control parameters with the large range of MRPI values between them.It was observed that control parameter D (gas flow rate) had the greatest impact on welded joint qualities, followed by control parameters A (current), B (speed) and C (voltage) in their orders as shown their ranks on table 7.

Analysis of variance
Analysis of variance (ANOVA) was used to determine either a given welding parameter had a significant influence on the quality characteristic or not.It demonstrated the effect of parameter on the weld quality and its significance on the process [2,16].As a result, the ANOVA was used to verify the relative effect of the control factors on the MRPI.In the ANOVA, the F-value and P-value of table 8 showed which variables had a substantial impact on the interested response.It was performed using a 95% confidence level and a significance threshold of 5%.If 0.05 is greater than factors P-value, it is presumed that the factor has a substantial influence on the output response, but if 0.05 is less than the P-value, the factor has no effect on the output response.Besides, if the F-value of parameter is greater than the F-critical value for the selected level of confidence, then the parameter is considered as significant unless it is considered as insignificant.
Factor C (volatge) was treated as a pooled factor in ANOVA to prevent zero degrees of freedom, due to its lowest significant influence as shown in table 8 [36].Current and gas flow rate were significant factors based on the ANOVA results of the TIG welding because their P-values were less than 0.05 and F-values were greater than the critical F-value, 19; however, the welding speed was insignificance since its P-value was greater than 0.05 and Where, ø m is total mean of MRPI, ø s is optimal levels' mean of MRPI for significant factor and k is number of significant control factors.
Equation (3) was used to compute an interval of confidence ♡ CI for the anticipated mean on the test of confirmation at 95% confidence value [2,16,25].
Where, a( ) F f 1, e is = 18.51 that was obtained from standard F ratio table, risk α = 0.05, f e is pooled errors degree of freedom (dof) = 2, V e is mean square of pooled error = 0.000934 as shown on table 8, N is total number of experiment = 9, S dof is total dof of significant factors = 4, n e is replication effective number = N/(1 + S dof ) = (9/(1 + 4)) = 1.8, and r is number of repetitions for confirmation experiment = 5.
The interval of 95% confidence for the predicted mean of the MRPI at the optimal parameter setting on the test of confirmation was found 0.6171 ± ♡ CI = 0.6171 ± 0.1143 = [0.5028,0.7314].

Validation experiments
Validation experiments were conducted to confirm the obtained optimal levels of the TIG welding with five repetitions or confirmation experiments at the optimal condition, A 3 B 1 C 2 D 2 .The confirmation tests MRPI mean value was obtained 0.6068, which was within the 95% confidence interval of 0.5028 and 0.7314.And, average values of the ultimate tensile strength, bending strength and Rockwell hardness of the confirmation tests were 614.8 MPa, 765.32 MPa, and 95.3 HRB, respectively.The MRPI values of both the confirmation tests and the predicted values are shown in table 9, and the experiment was safe.

Conclusion
The 304 austenitic stainless steel sheet metal is the most commonly utilized stainless steel grade in the world, and it is used almost in every sectors and industries, such as food, chemicals, textiles, aerospace and transportation.However, research was not carried out on 2 mm thickness of 304 austenitic stainless steel sheet metal to optimize the TIG welding parameters for multiple responses simultaneously.As a result, this study was conducted to fill the observed research gap by optimizing TIG welding parameters using the fuzzy logic based Taguchi method, and the following conclusions were drawn.
• The developed automated TIG welding fixture helped the welding torch to move at a constant speed, and it reduced the possible inherited defects from the variable welding speed of human.
• For the experimental trials of TIG welding process on 2 mm thickness of 304 austenitic stainless steel sheet metal, four welding parameters of speed, current, voltage and gas flow rate with three levels were used.As a result, nine experiments were conducted using L9 orthogonal array of Taguchi design.And then, experimental results of bending strength (BS), ultimate tensile strength (UTS) and Rockwell hardness (HR) of the butt weld joint were measured.In this experimental study, the highest values of the UTS of 615.7 MPa, BS of 765.4 MPa and HR of 94.3 HRB were obtained, and the lowest of values of the UTS, BS and HR were 505.9 MPa, 639 MPa and 88.25 HRB, respectively.
• The types of fractures and their locations were inspected visually on tensile tested specimens, and it was found that all specimens showed brittle fracture type and their fractures' locations were at fusion zones.
• The UTS, BS and HR results of the automated TIG welded 304 austenitic stainless steel sheet metal were predicted and optimized using the fuzzy logic based Taguchi method.The optimal multi-response performance index (MRPI) values were obtained with the optimal levels of parameters at 110 A of current, 13.86 cm/min of speed, 17.5 V of voltage, and 7.5 L/min of gas flow rate.
• According to the ANOVA for MRPI, the contributions of current, speed, gas flow rate and error were obtained 34.34%, 16.49%, 47.63% and 1.54%, respectively.Based on the F-value and P-value for 95% confidence level, current and gas flow rate were found as the significant factors.
• Finally, five confirmation experiments were conducted using the optimal levels of TIG welding parameters to verify the optimization.From the confirmation experiment, the average values of 614.8 MPa, 765.32 MPa and 95.3 HRB of UTS, BS and HR were obtained, respectively.The mean value of MRPI of the confirmatory test was 0.6068, and it was found between the 95% confidence interval of 0.5028 and 0.7314 which indicated that the experiment was safe.

Figure 1 .
Figure 1.Experimental setup of the TIG welding process.

Figure 2 .
Figure 2. Pareto chart for welding parameters (C=current, G=gas flow rate, V=voltage, S=speed, and RG=root gap for this graph only).

Figure 6 .
Figure 6.Optimization flowchart of fuzzy logic based Taguchi method.

Figure 11 .
Figure 11.UTS, BS and HR versus gas flow rate.

Figure 12 .
Figure 12.Determination of MRPI values in fuzzy logic.

Figure 13 .
Figure 13.FIS Editor with three inputs and one output.

Figure 14 .
Figure 14.Ultimate tensile strength input variable membership function.

Figure 19 .
Figure 19.Structure of the developed model.

Figure 20 .
Figure 20.Inputs and outputs from the fuzzy rule-based system.

Figure 21 .
Figure 21.Comparison of fuzzy logic and multi linear regression models.

Table 1 .
Chemical composition of AISI 304 austenitic SS sheet metal.

Table 2 .
Values of the control TIG welding parameters.

Table 3 .
Values of the fixed TIG welding parameters.

Table 5 .
Fracture types and their locations for tensile tested specimens.
a CR = Cooling rate.b HI = Heat input.

Table 6 .
MRPI values for S/N ratios values of UTS, BS and HR.
F-value was less than the critical F-value, 19.The contributions of gas flow rate, current, speed and error were 47.63%, 34.34%, 16.49% and 1.54%, respectively.The relevance of control factors was also assessed as D > A > B > C.
3.7.Confirmation tests 3.7.1.Predicting optimal mean valueA confirmation experiments purpose is to confirm the ideal circumstances for reducing variation.If the predicted and observed MRPI values of the several performance metrics are close to each other, the effectiveness

Table 7 .
Response table of MRPI values.
a Indicated the pooled process parameter.