Effect of Friction Coefficient in Friction Stir Welding of B4C Reinforced AA5083 Metal Matrix Composites and Use of Fuzzy Clustering Technique for Weld Strength Prediction

Te friction stir welding (FSW) method was used to weld B4C reinforced AA 5083 metal matrix composites in this study. By coating titanium nitride (TiN), aluminium chromium nitride (AlCrN), and diamond-like carbon (DLC) to a thickness of 4 microns, three FSW tools with square pin profles were developed and the friction coefcients of 0.69, 0.32, and 0.2 were maintained. At three levels, the process factors such as tool rotating speed, transverse feed, and axial force were examined. For each tool, 15 samples were made using the central composite design. Te infuence of the friction coefcient on ultimate tensile strength, microstructural features, and tool condition was studied, and the fower pollination algorithm (FPA) technique was used to fnd the best process parameters for obtaining maximum ultimate tensile strength of FSW joints. Te improved tensile strength of FSW joints was verifed using a validation test. Te coating has a considerable infuence on the ultimate tensile strength, microstructure, and tool condition, according to the results of the tool’s friction coefcient. Te results on the prediction of strength using the fuzzy clustering technique showed that the technique is efective in predicting the tensile strength values, with the root mean square error (RSME) of TiN, AlCrN, and DLC being 0.0027, 0.0016, and 0.0015, respectively, and the low RSME indicating that the prediction based on the fuzzy subtractive clustering technique is perfect and efective.


Introduction
In order to meet good fuel economy, automobile manufacturers are increasingly moving towards metal matrix composites (MMCs).Most of the models in modern automobiles including Lotus Elise, General Motors EVI, Chrysler Prowler, Volkswagen, and Toyota have used metal matrix composites for manufacturing their automobile parts [1].Te composite's desired qualities are achieved by combining the reinforcement's strength with the ductility of the matrix material.Despite the fact that MMCs have existed since the 1960s, the materials have yet to be fully commercialized due to their higher manufacturing costs and a lack of knowledge of the material's performance at extreme temperatures [2].Fusion joining, solid-state joining, and other similar methods are used to join the composite materials.Tere are a few challenges that must be overcome during the fusion welding of composite materials, particularly particulate-reinforced MMCs, such as melting above the melting point, the high viscosity of the melt, matrix, and reinforcement interaction, gas evolution, and segregation during solidifcation.Te friction stir welding (FSW) process has shown potential benefts in minimzing the fusion welding related problems of aluminium alloys.Gopalakrishnan et al. [3] used a mathematical model that includes welding factors such as axial force, tool revolving speed, traverse feed, additional reinforcement %, and tool pin shape to predict the tensile strength of the specimen connected.In his research, TiC particles are used to reinforce the AA 6061 material.Te efciency of the joint achieved in the majority of the joints created was greater than 90%.Te profle of the tool pin was shown to be more dominant than other characteristics in the established model.Several studies have recently attempted FSW process for joining various metal and composite materials [4][5][6].Parikh et al. [7] attempted to examine the articles on the FSW of metal matrix composites that were accessible.His thorough examination found that research including multiobjective optimization of process parameters in FSW will improve the utilization of composites in manufacturing.Tool wear, particularly the pin section of the tool, is still a major issue that must be addressed.
Hassen et al. investigated the wear behaviour of the welded connection while welding hybrid composites using various process parameters and pin profles.Te square pin tool had better wear resistance, which increased as the traverse feed was increased and the rotating speed was reduced in the experiments [8].Various base materials, such as AA6061, AA7005, and AA2124, were strengthened with Al 2 O 3 , SiC, TiB 2 , B 4 C, and ZrB 2 , and joined by friction stir welding [9][10][11][12][13][14].
Ali et al. [15] focused on studying the process parameter efect on properties of friction stir welded AL 6061 metal matrix composites reinforced with diferent SiC and B 4 C particles.Te test results showed that the maximum tensile strength of value 174 MPa was obtained for 10% SiC and 3% B 4 C composites.Te percentage elongation increases with increasing B 4 C and decreases with reducing the SiC components.Te wear resistance increases with the addition of reinforcement in Aal-metal matrix composites and the wear rate increases with an increase in load.
Prasad et al. [16] made an attempt to join 5 mm thick AA6061 T6 and AA6351-T6 aluminium dissimilar plates.Te welded surfaces were free from cracks and the maximum tensile strength of 167.95 MPa was reported for the square tool pin.Yadav et al. [17] performed an experimental and numerical study in the thermal feld and characteristics of the weld during submerged arc welding.Te model controls the melting and solidifcation phases during welding.Te microstructural investigation was also carried out in which the various distinct zones were clearly visible.Hasan et al. [18] fabricated copper alloy Bronze RG 10 reinforced with TiC and graphite to improve the wear behaviour by powder metallurgy route.Te results showed signifcant decrease in the wear rate when increasing the percentage of TiC and graphite.Te composites were best suited for coupling parts in the aerospace and automobile sectors.
Rao et al. [19] focused on the aging assessment of transformer insulation oil based on the integrated design of a fuzzy clustering system and found that the methodology adopted was efective and efcient in classifying the age group of the oil.Ali et al. [20] proposed the use of a subtractive fuzzy clustering technique for modeling soil cation exchange capacity and the results showed that the fuzzy subtractive clustering algorithm had high accuracy in predicting and modeling soil ratio exchange capacity.Yazdani Chamzini et al. [21] used the fuzzy model-based subtractive clustering technique to predict the performance of road header.Te Takagi-Sugeno (TS) fuzzy system based on the subtractive clustering method was efective and capable in fnding the complex relationship between road header and other parameters.
Pathak and Adhyaru [22] used model reference adaptive control based fuzzy subtractive clustering method to control the motion parameters of DC servo motor.Te results showed that FSC is more efcient and efective in reading the chattering and error convergence.It was observed from the literature that the fuzzy clustering technique was less applied in predicting mechanical properties.In the present work, an attempt was made to apply the fuzzy subtractive clustering technique to predict the tensile strength of the welded specimen.
Te B 4 C-reinforced aluminium MMCsare widely used in bullet-proof vests, bicycle frames, armor tanks, nuclear waste containment, neutron absorbers in nuclear power plants, and many transportation applications.Te use of the material in these applications requires the joining with the same or diferent materials using diferent joining methods.From the literature study, it was noticed that only limited work has been reported in the joining of B4C-reinforcedAA 5083 MMCs.Te novelty of this research article resides in analyzing the weld characteristics in terms of efect of friction coefcient and use of the fower pollination algorithm (FPA) technique to fnd the best process parameters for maximizing and predictingthe ultimate tensile strength of friction stir welded B4C reinforced AA5083 MMC joints using the fuzzy clustering technique.
Te main objective of this research article is to study the infuence of the friction coefcient on ultimate tensile strength and microstructural features of FSW joints and analyzing the FSW tool condition after welding.Te fower pollination algorithm (FPA) technique was used to fnd the best process parameters for obtaining maximum ultimate tensile strength of FSW joints.
In this proposed research work, the AA5083 alloy reinforced with 10% volume fraction of boron carbide MMCs were made using liquid stir casting process with dimensions of 10 cm × 5 cm × 0.6 cm.Using a CNC milling machine, the composites are machined to the desired specifcations.Te produced composites were joined using a solid state friction stir welding (FSW) procedure with three distinct ranges of tool rotational speed, traverse feed, and axial force.Tree tools with square pin profle were fabricated andcoated with TiN, AlCrN, and DLC to a thickness of four microns for changing the friction coefcient and wear resistance of FSW tool.
Figure 1 depicts the suggested work plan as a fow chart.In the experimental Sections 2.1 and 2.2, the specifcs of the process parameter selection, tool design and coating, composite preparation, and experimentation are described.Part 1 discussion is ofered as 3.1, 3.2, and 3.3 in the fndings and discussion section.Te observed UTS values were used to train fuzzy clustering to predict the strength which was explained in Section 3.4.

Selection of Materials.
Te basis material was nonheat treatable medium strength aluminium AA 5083, which has uses in crucial arctic, maritime structures, storage, and transportation tanks for cryogenic fuids.Te stir casting 2 Advances in Materials Science and Engineering process was used to manufacture composite blocks reinforced with a boron carbide particle with an average particle size of 10 microns in AA 5083 to a 10% volume fraction.
Liquid stir casting is one of the most appropriate technologies for producing composites with a volume fraction of up to 30%.A crucible was used to melt the AA 5083 aluminium alloy in an electrical furnace.To make metal matrix composites, weighed amounts of B 4 C particles were added to molten aluminium metal.To guarantee equal particle dispersion in the molten metal, a mechanical stirrer with an electric motor was used to agitate the melt.Te temperature of the molten metal was kept at 850 °C during the composites' preparation, and it was placed into a prepared die and allowed to harden.Te die was made of hot wrought steel and was 10.5 cm × 10.5 cm × 0.8 cm in size.Te composite blocks were visually examined for surface faws before being cut to a size of 10.4 cm × 5 cm × 0.6 cm using a vertical milling machine.Figure 2 depicts the sample composite blocks.Te tool is a critical component in creating a defect-free weld in friction stir welding.Te spinning tool's primary duty is to heat the base material and to allow it to fow in both horizontal and vertical directions.It was clear from the literature that pins with a square shape provided defect-free welding.Te tool material's friction coefcient has an efect on heat generation and material fow characteristics.
Tree tools were machined and coated with TiN, AlCrN, and DLC to change their hardness and friction coefcient.Te basic characteristic of the coating is given in Table 1.

Conducting the Experiments.
Rotational speed, traverse feed, and axial force were used as process variables in this study and were measured in three distinct ranges.Table 2 lists the process parameters and their ranges.
Te number of experiments was determined using the formula 2 k + 2K + C, where C is the number of central points and K is the number of factors considered [23].Using a computer-controlled FSW machine, a total of 15 trials were carried out for each coated tool.

Ultimate Tensile Strength (UTS).
Tensile strength is one of the most important properties for determining a material's mechanical performance, particularly for freshly created materials such as composites.Te welded workpieces were visually inspected for surface problems before samples were machined using the wire EDM method to test the ultimate tensile strength according to ASTM E8 standards with 32 mm gauge widths of 6 mm gauge lengths and overall lengths of 100 mm.Te specimens were put through their paces in a computerized UTM with a load capacity of 100 kN. Figure 3 depicts the samples of welded joint's tensile test specimen joined using the AlCrN-coated tool after the test.
From the tested specimens, it was noticed that specimens joined using the AlCrN-coated tool failed at the heat afected zone, whereas the specimens joined using the DLC-coated tool failed at the stir zone which indicates that stirring was not proper and due to which tensile strength was afected.It was observed from the test results that the specimens which got a break at the heat afected zone had resulted in better tensile strength and the components that failed at the stir zone had resulted in lower tensileAlCrN coated.
Te AlCrN-coated tool had the highest ultimate tensile strength of 153 MPa, followed by the TiN-coated tool with 136 MPa and the DLC-coated tool with 130 MPa, according to the experimental data.Figure 4 depicts the ultimate tensile strength fndings.Te experimental results demonstrated that the fuctuation of UTS was minimal for all the three coated tools, with a signifcant diference between them.Tis implies that the discrepancy is attributable to a diference in the tool coating friction coefcient.
Design expert 12.0 version software was used to create the mathematical model for all the three coated tools.For TiN-, AlCrN-, and DLC-coated tools, the model equation in the coded form is given in equations ( 1)-(3), respectively.
In this investigation, the F ratio for the titanium nitratecoated tool is 19.60, with a 0.0496 discrepancy between anticipated and adjusted R 2 .Te P value of the TiN-coated tool is 0.0022.Te diamond-like carbon-coated tool has an F ratio of 35.58, a diference of 0.0276, and a P value of 0.0005, whereas the aluminium chromium nitrate-coated  Advances in Materials Science and Engineering tool has an F ratio of 11.39, a diference of 0.064, and a P value of 0.0077.Te ANOVA test results for the tensile strength of all three tools are shown in Table 3.

Optimization Using the Flower Pollination Algorithm.
Te fower pollination algorithm is a nature-inspired metaheuristic algorithm that replicates pollination behaviour in blossoming plants.Yang developed this approach in 2012, and it outperforms other metaheuristic algorithms [24].Te fower pollination approach was utilized in this research to get the best ultimate tensile strength values.Te MATLAB R 2018a version program was used to estimate the ideal ultimate tensile strength and its associated process parameters.
A confrmation test was performed on the improved settings, with the results compared to the expected values.

Microstructural Investigation.
Te microstructural analysis was carried out in this study for the joints that provided the highest and least tensile strength with all three coated tools.Te maximum strength was 136 MPa for the process parameters of rotational speed 1500 rpm, traverse feed 40 mm/min, and axial force 7 kN, according to the fndings of the tensile test.Te welded component joined by the TiN-coated tool was examined under a microscope with a magnifcation range of 5x-300,000x and an accelerating voltage of 0.3-30 Kv, using a Hitachi SU1510 model with a magnifcation range of 5x-300,000x.
Te welded specimen's base material is shown in Figure 5(a).Te boron carbide particles were uniformly dispersed, as seen in the diagram.A small aggregation of particles was observed in a few sites.Te diferent zones of friction stir welding have been seen in Figures 5(b)-5(d).
Te grains were elongated and deformed patches were detected in the agitation zone of the specimen due to plastic deformation, as illustrated in Figure 6(c).Te process settings of 1800 rpm, 20 mm/min, and 7 KN resulted in unsatisfactory strength, as illustrated in Figures 5(e

) and 5(f ).
Tis happened because the TiN-coated tool had a lower hardness and a greater friction coefcient.Te amount of heat generated during the process for this combination was high due to the higher friction coefcient, higher rotational speed, and lower traverse feed, causing the aluminium particles to become stuck on the tool materials' surface, resulting in small voids and tunnels in the weld zone, as shown in Figures 5(e) and 5(f ).
Figures 6(a)-6(d) show the optical images taken corresponding to the specimen which produced the maximum strength of 153 MPa for 1500 rpm, 40 mm/min, and 7 kN using the AlCrN-coated tool.During the process, stirring was efective due to the nominal value of friction coefcient and hardness.Te distinct regions such as stir zone, TMAZ, and HAZ are clearly shown.Te onion rings were formed near the stir zone due to the frictional heat generated by the    7(a)-7(c) show that the stirring of the material is not proper and shiny and skinny surfaces were observed due to the slipping of the tool during stirring.Figures 7(e)−7(f) show the optical microscopic images of the specimen prepared for 1500 rpm, 40 mm/min, and 7 kN.Te tensile strength measured for aforesaid parameters was 129 MPa.In the optical images, the various weld regions such as stir zone, heat afeted zone, and thermomechanically afected zone were observed clearly.

Condition of the Tool.
Te FSW tool was subjected to one of the following three circumstances during the stirring process: maximum sliding, maximum sticking, and partial sliding and/ or partial sticking.Images collected with the use of videomeasuring equipment reveal that the aluminium particles became adhered to the surface of the TiN coated tool in the current experiment, as shown in Figures 8(a)-8(c).During the process, when the yield strength of the workpiece is less than shear stress produced in the contact area.Small segments of workpiece materials would stick to the tool surface which is considered as phenomenon called more sticking.Te friction coefcient of the tool material greatly controls the amount of shear stress developed.Normally, a higher friction coefcient leads to the higher shear stress.In Figures 8(d)-8(f), it was noticed that white-colored materials were stick on side surface of the tool.
Te pin of the DLC-coated tool, on the other hand, had completely worn out throughout the operation.Te DLCcoated tool has a friction coefcient of roughly 0.2, resulting in a shear stress that is lower than the material's yield strength.Te extrusion of the materials was carried out only by the pin in this condition of stirring, which is referred to as complete sliding.Te pinned piece has completely worn away as a result of these circumstances.
Figure 9 shows an optical microscopic picture acquired to better understand the surface behaviour of the worn tool.Te photograph was taken on the shoulder's upper surface, which is in touch with the workpiece materials.It has parallel grooves, indicating that it was worn during the manufacturing process.

Prediction Based on Fuzzy
Clustering.Soft computing approaches may be used to fnd solutions to complicated systems that are nonlinear, continuous, discrete, or dynamic System modeling may be done with fuzzy sets, which are sets having inaccurate or hazy bounds.Te two major classes involved in the application fuzzy set known as fuzzy logic controller are linguistic fuzzy modeling and precise fuzzy modeling (FLC).Clustering, a strong data mining approach, is a powerful tool for extracting usable information from a set of data.After examining the dataset, the clustering algorithm divides the data into numerous clusters based on pattern similarities [32].
Tere are three primary processes in the fuzzy inference system based on Chiu's fuzzy subtractive clustering approach.In an M-dimensional space, for example, with data   3) is used to calculate the potential of a p i data point.
Te neighborhood's range is controlled by the value r a , which is a positive constant.After computing all of the potentials for each data point, the cluster centre with the highest potential is chosen as the initial cluster centre.Te second step is to use equation ( 5) to revise the potential for the available data points other than the designated cluster centre.Te cluster centre is determined after the largest potential value point.Te neighborhood range is controlled by the constant r b , which is set at 1.5 times r a .(5) Finally, the preceding two processes are repeated iteratively until all of the data falls inside the threshold range (∈) � 0.5 to any cluster centre.Fuzzy subtractive clustering is implemented in MATLAB using the "genfs" options.Te input membership function for each fuzzy cluster is in Gaussian format, and one rule is constructed for each fuzzy cluster based on subtractive clustering.Based on a precise fuzzy modeling method, a fuzzy inference system developed using "genfs" options may predict the output parameter.As a confrmation test, the anticipated values were compared to the optimized values, and experiments were undertaken for the optimized parameters.Table 4 shows the outcomes of the tests compared.
When the predicted values were compared to the experimental values, the root mean square error (RSME) of TiN, AlCrN, and DLC was found to be 0.0027, 0.0016, and 0.0015, respectively, indicating that the UTS prediction based on fuzzy subtractive clustering approach is excellent and efective.

Conclusion
Using three coated tools with varying friction coefcients and hardness values, an efort was made to join B4Creinforced AA 5083 metal matrix composites by friction stir welding.Te following observations were drawn from the experimental investigation.
(i) Among the three tools, the AlCrN-coated tool produced the maximum tensile strength of 153 MPa followed by the TiN-and DLC-coated tools at 136 MPa and 130 MPa, respectively.(ii) Even though process parameters were changed at three various levels, for the same tool, the tensile strength was not changed drastically.On contrary, among the three tools, a noticeable variation was observed and this variation was due to the change in the friction coefcient of the tool coating.(iii) Te fower pollination algorithm (FPA) was used to fnd the optimized process parameters, and the results were tested by doing a confrmation test on the optimized process parameters.Tese test results were also compared to the predicted values, and the percentage variation was found to be less than 5%.Te results proved that the optimization method and prediction models were accurate.(iv) Te microstructural analysis of weldments prepared by the TiN-coated tool showed small agglomeration of reinforcement particles.Te stir zone of the specimen had elongated grains due to plastic deformation.At a larger friction coefcient, higher rotational speed, and lower traverse feed, more amount of heat was developed and aluminium particles got stuck on the tool surface which was refected as small tunnels and voids.Due to the nominal friction coefcient, the AlCrN-coated tool induced efective stirring.Te optical microstructure showed clear distinct regions and observed onion rings near the stir zone.(v) For the DLC-coated tool, due to the minimum friction coefcient, the amount of heat developed was less, which afected the stirring of the process.(vi) Te condition of the tool was analyzed by the images taken using the video measuring system.Te TiN-coated tool undergoes maximum sticking phenomenon in which the aluminium particles got stuck to the surface of the tool pin due to the higher friction coefcient.Te results revealed that the AlCrN-coated tool was subjected to partial sliding and partial sticking.Te DLC-coated tool underwent full sliding where the extrusion of the material was carried fully by the tool pin which resulted in tool wear.(vii) Te results of the fuzzy clustering showed that the technique was efectively used in predicting the

Figure 1 :
Figure 1: Plan of the proposed work as a fow chart.

Figure 3 :
Figure 3: Sample tensile test specimen after the test.

Figure 4 :
Figure 4: Comparison of UTS among three tools.

Figure 9 :
Figure 9: Worn surface of the DLC tool.
p i � p i − p i * e −β ‖p i − p j ‖ 2 ,

Table 1 :
Characteristics of the coating.

Table 2 :
Process parameters and their range.
Reproduced from Devanathan et al., 2021, under the creative commons attribution license/public domain.

Table 3 :
ANOVA test results of UTS.

Table 4 :
Comparison of test results.Advances in Materials Science and Engineering ultimate tensile strength with the root mean square error (RMSE) values of 0.0027, 0.0016, and 0.0015 for TiN, AlCrN, and DLC coated tools, respectively.Te minimum RMSE values indicated that the clustering technique can be efectively used to predict the ultimate tensile strength for any parameters within the range without conducting the experiments to save money and time.