Sage Journals: Discover world-class research

Abstract

Friction stir welding of lightweight aluminium alloys have advantage in automobile industry with its vast applications. This research work focuses on the influence of FSW process parameters on novel interlock lap weld of AA7075-T7-AA7475 tailor welded blank. Three levels of the parameters, including tool rotation speed (TRS), weld speed (WS) and plunge speed, were used to form L27 orthogonal array to optimize the input process conditions. Ultimate tensile strength and Vicker's micro hardness were measured to test the characteristics of the interlock welded samples. Scanning electron microscopy analyses have been carried out to study the surface morphologies and elemental components in the welded samples. Artificial neural network (ANN) has been used to predict the optimized process parameter associated with the novel interlock lap weld. The TRS and WS contributed significantly in improving the mechanical behaviour and microstructural characteristics of interlock lap welds. Visual inspection and surface morphology analysis showed uniform dispersal of aluminium alloy deposition throughout the interlock weld samples. The ultimate tensile strength and micro hardness prediction was carried out using ANN with 95% accuracy level. The predicted results of ANN were more accurate than the experimental results and regression model of fractional factorial design. The defined FSW interlock lap weld stands out as the substitute for typical FSW lap weld of aluminium alloys which fulfils the modern automotive industry demands in welding monocock frames.

Keywords

Interlock FSW full factorial optimization artificial neural network SEM

Introduction

The utilization of different grade material structures in growing modern applications can effectively trade-off between the cost, weight and performance in transportation and other industries.¹ Aluminium alloys of various grade materials have been utilized at most with various grades of other light weight metals, as single metal part or tailor welded blank and their combination can improve the functionality of the light weight structures irrespective of component size utilized. Friction stir welding (FSW) remains one of the prominent and versatile technique to join dissimilar metals and alloys in order to meet the demands of the modern production.² In conventional welding technique, the uncontrollable parameters which influence the weld quality are excessive heat generation, surface cracks, porosity, pits and voids occurring on the nugget surface of the welds. FSW, a non-conventional technique, has a potential to limit the occurrence of these defects. This technique combines the effect of non-consumable rotating tool and force.³ In FSW, the plasticization of alloys above its melting point improves the statistical and microstructural properties of the weld. The controllable input variables TRS, WS, PSR, plunge depth, dwell time and tool tilt angle influence the quality of welding.⁴ Three important stages in the FSW are plunging, dwelling and welding. Improper selection of input process parameters may lead to production of defective quality welds.⁵ Kadlec et al. investigated the mechanical property of AA7475 with various surface defects and, suggested that AA7475 alloys have better mixing ability with other alloys and improved the hardness and tensile strength to a greater extent.⁶ Dissimilar aluminium alloys fabricated with higher tool rotation speeds (TRSs) of 1100, 1400 rpm has reported with increased tensile strength and hardness. Increasing the weld speed (WS) reduced the occurrence of weld defects. Improper selection plunge speed rate decreased the strength of the joints and lead to the formation of keyhole defects and, dwell time did not show any significant contribution.⁷ 7xxx series aluminium alloys generally possess higher strength, %elongation and shear hardness which is suitable for the welding of monocoque structures in automobile, for example, 7xxx series aluminium alloys are commonly used in aircraft applications for their higher rigidity and their ability to sustain at higher impact force without any deformation.⁸ AA 7075 is heat treatable and exhibits excellent properties such as ductility, good resistance to fatigue. AA7475 exhibits excellent corrosion resistance in addition to the mechanical attributes such as yield strength, fatigue strength and tensile strength. Interlock FSW joints significantly improves the strength of the lap joints.⁹ A research on refill FSSW of AA 7475-T761 was conducted by Kwee et al. and made clear that the link between plunge depth and dwell duration generated good welds with increased hardness and ultimate strength.¹⁰ Hence, this combination of 7xxx series aluminium alloys is preferred in automobile manufacturing. The literature review clearly shows that there is high potential among aluminium tailor welded blanks in industrial applications and very limited study is available in FSW interlock lap welds. In this research, the effect of shear tensile force and microstructural properties, material flow behaviour, intermixing, hardness distribution of FSW interlock dissimilar lap welds AA7075 T6-AA7475 T7 aluminium alloy has been investigated with full factorial technique. The uniqueness of selected input responses was enhanced using ANN technique. Hardness (HV) and tensile strength (TS) of interlock welded joints were investigated. Hence, the objective of this research work is to successfully weld and characterize the FSWed interlock lap joint.

Materials selection and preparation

The commercial aluminium alloys, AA7075 T6-AA7475 T7, are used to carry out the FSW process and create interlock twb's, as 7XXX series aluminium alloys are widely used in automobile applications due to its high strength and machinability. Also, these aluminium grades have excellent compressive strength and wear resistance. Table 1 shows the mechanical properties of AA7075 T6-AA7475 T7.

Table 1.

Mechanical properties of AA7075 T6-AA7475 T7.

Base alloys/wt%	Si	Ti	Zn	Mg	Fe	Mn	Cu	Cr	Al
AA7075-T6	0.12	0.03	5.9	2.3	0.15	0.5	2.3	0.03	88
AA7475-T7	0.04	0.06	5.7	2.3	0.12	0.06	1.5	0.22	90

The aluminium alloy AA7475 T7 sheets were milled of 0.5 mm depth, width of 30 mm from the centre of the sheet which is placed at the bottom. Table 2 depicts the process parameters and their levels. The sheet which is placed on top AA 7075 T6 is also milled for a depth of 0.5 mm, breadth of 20 mm such that it interferes with unmilled part of the bottom AA 7475 T7 to form an interlock between the joints.¹¹ The graphical representations of the interlock FSW lap joints are shown in Figure 1.

Figure 1.

Graphical representation of interlock lap joint.

Table 2.

Process parameters and their levels.

Factors	Symbol	Unit	Level 1	Level 2	Level 3
Tool rotation speed	TRS	rpm	1000	1200	1400
Weld speed	WS	mm/min	24	30	36
Plunge speed rate	PSR	mm/s	0.04	0.06	0.08

Full factorial design of experiments

The complex conventional investigational plan is not feasible as it involves many iterations. Full factorial technique employs less number of experiments with greater accuracy.¹² The mechanical properties of welded samples are indicated by analysis of variance and strength-to-noise (S/N) ratio.¹³

Statistical characterizations

Figure 2(a) shows the universal hardness tester to determine the hardness of welded samples. The sample has been prepared to ASTM E8 standard to measure the ultimate tensile strength¹⁴ (Talk about the adjustment used to avoid shear.). UTS of the welded samples is the resistance to break under tension when external load is applied.¹⁵ The diamond ball indenter was exposed to a load of 0.5 kgf for 15 s at 15 different positions and, the average HV was taken.

Figure 2.

(a) Hardness tester, (b) Shimadzu Tensile Machine, with specimen loaded.

Results

Statistical analysis

L27 orthogonal array was carried out to study the statistical properties with respect to quality, strength and defect free characteristics opted from the input process parameters considered for carrying out the experiments. Tensile strength and hardness are the measured characteristic property based on the FSW input process parameters.¹⁶ Minitab 18 has been used to carry out the L27 design of experiments with full factorial orthogonal array.¹⁷ “Larger is better” S/N ratio has been selected to identify the optimum FSW input parameters. Table 3 depicts the 27 experiments, signal-to-noise ratio for ultimate tensile strength and hardness. ANOVA and S/N ratio associated with the impact and optimum condition for the selected input factors. Figures 3, 4(a) and (b) depict the S/N ratio for hardness and tensile strength. The TRS 1400 rpm, WS 30 mm/s, PSR 0.06 mm/s are the optimum input parameters for tensile strength and hardness of the interlock weld samples. “Larger is the better” S/N ratio depicted the highest hardness and tensile strength. The influence of process parameters on tensile strength and hardness with respect to S/N ratio is represented in Tables 4 and 5. Optimum condition for obtaining maximized output is TRS3, WS2, PSR2 (Table 2), that is, TRS at level 3 (TRS3), WS at level 2 (WS2) and plunge speed rate at level 2 (PSR2).

Figure 3.

Response curve for tensile strength.

Figure 4.

Response curve for hardness.

Table 3.

Three-level three factor (3³) full factorial design of experiments.

S. no	Tool rotational speed (rpm)	Weld speed (mm/s)	Plunge speed rate (mm/s)	Ultimate tensile strength (MPa)	Yield strength (MPa)	Hardness (HV)	S/N ratio for tensile strength	S/N ratio for hardness
L1	1000	24	0.04	51.888	37.254	129	34.301	42.2118
L2	1000	24	0.06	97.173	72.2354	148	39.751	43.4052
L3	1000	24	0.08	90.355	68.1589	137	39.119	42.7344
L4	1000	30	0.04	80.712	60.3258	153	38.139	43.6938
L5	1000	30	0.06	105.68	75.2569	157	40.48	43.9180
L6	1000	30	0.08	82.165	60.548	146	39.294	43.2871
L7	1000	36	0.04	72.742	52.254	144	37.236	43.1672
L8	1000	36	0.06	107.68	75.5687	150	40.643	43.5218
L9	1000	36	0.08	75.529	55.369	149	37.562	43.4637
L10	1200	24	0.04	73.916	55.2563	158	37.375	43.9731
L11	1200	24	0.06	136.06	110.25	166	42.675	44.4022
L12	1200	24	0.08	93.008	70.754	161	39.37	44.1365
L13	1200	30	0.04	82.101	62.2154	163	38.287	44.2438
L14	1200	30	0.06	148.88	112.88	170	43.457	44.6090
L15	1200	30	0.08	84.231	62.89	164	38.509	44.2969
L16	1200	36	0.04	79.638	55.485	157	38.022	43.9180
L17	1200	36	0.06	123.95	105.75	162	41.865	44.1903
L18	1200	36	0.08	82.217	61.214	158	38.299	43.9731
L19	1400	24	0.04	91.081	70.235	166	39.189	44.4022
L20	1400	24	0.06	145.54	115.25	183	43.01	45.2490
L21	1400	24	0.08	141.42	110.48	168	43.26	44.5062
L22	1400	30	0.04	155.57	123.32	181	43.839	45.1536
L23	1400	30	0.06	172.88	130.89	200	44.755	46.0206
L24	1400	30	0.08	120.5	97.45	186	41.62	45.3903
L25	1400	36	0.04	96.472	75.32	170	39.688	44.6090
L26	1400	36	0.06	152.94	122.235	178	43.961	45.0084
L27	1400	36	0.08	105.9	88.325	177	40.498	44.9595

Table 4.

Signal-to-noise ratio for tensile strength (larger is better).

Level	Tool rotational speed (rpm)	Weld speed (mm/s)	Plunge speed rate (mm/s)
1	40.56	39.18	39.98
2	38.88	40.54	39.89
3	42.26	40.23	40.45
Delta	3.38	1.03	0.56
Rank	1	2	3

Table 5.

Signal-to-noise ratio for hardness (larger is better).

Level	Tool rotational speed (rpm)	Weld speed (mm/s)	Plunge speed rate (mm/s)
1	43.54	43.81	44.34
2	44.10	44.54	44.29
3	44.86	44.14	43.87
Delta	1.32	01.73	0.47
Rank	1	2	3

Analysis of variance

Analysis of variance was used to identify the influence level of each input process parameters on the interlock FSWed samples. F-test analysis clearly depicts the contribution of each individual parameter. Figure 5 depicts the contribution percentage of input process parameters to the statistical characteristics.¹⁸ The percentage contributions of the TRS, WS and PSR for hardness and tensile strength are depicted in Figure 5(a) and (b). In Table 6, the controllable factors for achieving higher micro hardness are F-value and contribution percentage. The contributions of TRS are 75.2% and 9.95% of WS followed by the PSR with 7.23% and 7.62% error. Hence, it is evident that TRS is the prominent and higher contributor for increased tensile strength of the welded samples using FSW. Similarly for hardness as represented in Table 7, TRS has crucial role to attain the maximized hardness; it contributed 39.46% and 38.51% of WS followed by the PSR with 4.60% and 17.43% error. From analysis of data it is evident that TRS and WS contribute maximum to the strength and hardness of the interlock lap weld joints.

Figure 5.

Contribution of (a) tensile strength, (b) hardness for increased strength of interlock FSW lap weld.

Table 6.

ANOVA for UTS of interlock lap joint.

Source	Contribution%	DOF	Adj. SS	Adj. MS	F-value	P-value
TRS	75.2	2	4867.9	2433.93	98.66	0.000
WS	9.95	2	640.7	320.04	12.97	0.000
PSR	7.23	2	467.9	233.93	9.48	0.001
Error	7.62	20	493.4	24.67
Total	100	26	6469.2

Table 7.

ANOVA for hardness of interlock lap joint.

Source	Contribution%	DOF	Adj. SS	Adj. MS	F-value	P-value
TRS	39.46	2	10,078	5039.2	22.63	0.000
WS	38.51	2	1169	584.3	2.62	0.097
PSR	4.60	2	9834	4916.9	22.08	0.000
Error	17.43	20	4453	222.7
Total	100	26	25,534

Regression equation

The relation between the TRS, WS, plunge speed rate and obtained static results such as ultimate tensile strength, hardness were co-related by regression equations (1) and (2).¹⁹ The regression equation (1) depicts that the input process parameters TRSs (A₁, A₂), WSs (B₁, B₃) and plunge speed rate (C₁, C₃) are negative.

\begin{aligned} The regression equation for Ultimate tensile strength \\ = & 105.56 - 20.68 A_{1}_1000 - 5 .12 A_{2}_1200 \\ + 25 .80 A_{3}_1400 - 3 .29 B_{1}_24 + 9 .18 B_{2}_30 \\ - 5 .89 B_{3}_36 - 18 .44 C_{1}_0 .04 + 26 .29 C_{2}_0 .06 \\ - 7 .85 C_{3}_0.08 \end{aligned}

(1)

\begin{aligned} The regression equation for Hardness : \\ 162.259 - 16 .37 A_{1}_1000 - 0 .15 A_{2}_1200 \\ + 16 .52 A_{3}_1400 4 .93 B_{1}_24 + 6 .63 B_{2}_30 \\ - 1 .70 B_{3}_36 - 3 .48 C_{1}_0 .04 + 5 .85 C_{2}_0 .06 \\ - 2 .37 C_{3}_0.08 \end{aligned}

(2)From the regression equation of the hardness (2), the co-factors associated with TRSs (A₁, A₂), WSs (B₁, B₃) and plunge speed rate (C₁, C₃) are negative. The positive indication in the input process conditions shows that the particular parameter has more impact on the statistical properties of material. The input process conditions with positive sign in the equation increase the value of statistical properties, and with the negative sign decreases the statistical properties of the FSWed interlock lap welds. For input parameters optimization, formula created with the aid of full factorial design is:

\begin{aligned} ŋ HV = ŋ + (mTR S_{3} - ŋ) + (mW S_{2} - ŋ) + (mPS R_{2} - ŋ) \dots \dots \end{aligned}

(3)where, ŋ is the response mean of S/N value, the optimum range of S/N value for TRS at level 3 is taken as mTRS3; mWS2 is the optimum S/N data for WS at level 2; the PSR optimal S/N response at level 2 is taken as mPSR2. Overall, the mean for S/N tensile strength value is ŋ = 44.16 (Table 2).

Now by employing in equation (3), we get

ŋ = 44.16 + (42.26 - 44.16) + (40.54 - 44.16) + (39.89 - 44.16) = 34.37

For the condition larger is better:

y_{o p t =}^{2} 10 \frac{ŋ o p t}{10}

y_{o p t}^{2} = 10^{34.37} /^{10} 180.21 \overset{}{\to} 180.21

The optimum regression ultimate tensile strength is 180.21 Mpa

Similarly for Hardness (HV) S/N mean value ŋ = 45.33

ŋ = 45.33 + (44.86 - 45.33) + (44.54 - 45.33) + (44.29 - 45.33) = 49.63

For the condition larger is better:

y_{o p t =}^{2} 10 \frac{ŋ o p t}{10}

y_{o p t}^{2} = 10^{49.63} /^{10} \overset{}{\to} 200.44

The optimum regression Hardness is 200.44 HV.

Artificial neural network

Artificial neural network (ANN) is a technique which compares and solves data set acquired from other statistical techniques. ANN involves three layers an input, output and selection of number of hidden layers according to the data set obtained. The first input layer is used to feed the data set and the output layer produces the final result. The hidden layers in the ANN process the raw input data set for processing. TRS, WS, plunge speed rate are input process parameters and ultimate tensile strength and micro hardness are output parameters.²⁰ Levenberg-Marquardt training algorithm combined with feed forward back propagation ANN technique was used to predict and analyze the tensile and hardness experimental results. Four neurons are there in two hidden layers selected to predict the output data set.¹⁹ ANN divides the output and input into three subordinate groups: training (60%), testing (20%) and validation (20%). Out of 27 experimented data set, 17 data was selected for training, 5 data for validation and 5 data for testing.

Five neurons correspond to the three inputs that are fed in the input layer of ANN; five neurons were specified in the two output layers and two neurons were designated in the hidden layer. The developed ANN for input and output data set was fed in 3−5-5−2 architecture (Figure 6). The quality of welded joints was evaluated in terms of correlation coefficient and error prediction percentage. Figure 7 depicts the detailed exhaustive ANN prediction flow diagram from feeding input data set to prediction of output results. The percentage (%) of predicted error is calculated as

% of predicted error = \frac{Experimental data - ANN data}{Experimental data} \times 100

Figure 6.

ANN architecture.

Figure 7.

ANN prediction flow chart.

Table 7 represents the training, testing and validation data sets with the percentage of predicted error. The predictive performance of ANN model with three input process parameters is represented in Figure 8(a) and (b). The ANN and regression outputs were represented in Table 8. The total coefficient correlation for all models obtained was 0.97416 (HV) and 0.98749 (TS) and mean of predicted error % (0.91426) was less than ± 3%. Hence, the correlation coefficient exhibited good relationship between predicted and experimental values as depicted in Figure 9(a) and (b). Therefore, FSW input process parameters TRS, WS and PSR correlated with the ANN(3-5-5-2) model is used to predict the statistical properties such as tensile strength, hardness of interlock FSWed lap joint AA7075-T7-AA7475 with least percentage of error between the predicted and the experimental results. The full factorial experimental design and ANN results are compared in Figure 9(a) and (b). Verification of experimental results with ANN has been done when the parameters attain the desired optimum level. Tables 8 and 9 represent the comparison between ANN predicted data and experimental results. Therefore, it is evident that both full factorial design and ANN give near accurate results.

Figure 8.

Predictive performance of ANN model: (a) micro hardness predictive performance, (b) ultimate tensile strength predictive performance.

Figure 9.

Comparison of (a) ultimate tensile strength comparison, (b) hardness.

Table 8.

Predicted error percentage using ANN.

Trial no.	Experimental results		Predicted results		Predicted error (%)
Trial no.	UTS (MPa)	Hardness (HV)	UTS (MPa)	Hardness (HV)	UTS (MPa)	Hardness (HV)
1	51.888	129	50.84	130.3	2.02	−1.02
2	97.173	148	96.86	151.9	0.32	−2.62
3	90.355	137	90.65	135.4	−0.33	1.17
4	80.712	153	80.84	155.0	−0.16	−1.29
5	105.68	157	103.8	155.3	1.71	1.08
6	82.165	146	83.74	150.1	−1.92	−2.78
7	72.742	144	73.16	146.4	−0.57	−1.67
8	107.68	150	105.4	152.3	5.20	−1.53
9	75.529	149	74.18	155.7	1.79	−4.47
10	73.916	158	74.25	157.2	−0.45	0.51
11	136.06	166	133.2	169.5	2.05	−2.11
12	93.008	161	95.86	165.6	−3.07	−2.84
13	82.101	163	85.14	160.5	−3.70	1.53
14	148.88	170	146.4	173.4	1.65	−2.00
15	84.231	164	81.40	164.8	3.36	−0.49
16	79.638	157	78.50	154.2	1.43	1.78
17	123.95	162	129.4	167.4	−4.44	−3.33
18	82.217	158	81.54	159.2	0.82	−0.76
19	91.081	166	92.74	169.8	−1.82	−2.29
20	145.54	183	153.6	179.8	−5.57	1.75
21	141.42	168	143.5	179.1	−1.47	−6.62
22	155.57	181	158.3	184.2	−1.77	−1.77
23	172.88	200	170.8	197.9	1.18	1.07
24	120.55	186	124.1	194.7	−2.95	−4.69
25	96.472	170	95.72	173.2	0.78	−1.88
26	152.94	178	154.3	179.5	−0.90	−0.84
27	105.95	177	107.3	173.7	−1.29	1.86

Table 9.

Regression and ANN experimental results.

Trial	TRS (rpm)	WS (mm/s)	PSR (mm/s)	Ultimate tensile strength (MPa)	Hardness (HV)	Regression output tensile strength	Regression output hardness	ANN output TS	ANN output HV
L1	1000	24	0.04	51.888	129	60.258	129.69	50.84	130.3
L2	1000	24	0.06	97.173	148	105.23	148.96	96.86	151.9
L3	1000	24	0.08	90.355	137	98.521	137.19	90.65	135.4
L4	1000	30	0.04	80.712	153	83.654	153.44	80.84	155.0
L5	1000	30	0.06	105.68	157	113.74	157.85	103.8	155.3
L6	1000	30	0.08	82.165	146	83.215	146.32	83.74	150.1
L7	1000	36	0.04	72.742	144	75.564	144.69	73.16	146.4
L8	1000	36	0.06	107.68	150	120.47	150.44	105.4	152.3
L9	1000	36	0.08	75.529	149	78.329	149.19	74.18	155.7
L10	1200	24	0.04	73.916	158	84.548	158.55	74.25	157.2
L11	1200	24	0.06	136.06	166	149.23	166.69	133.2	169.5
L12	1200	24	0.08	93.008	161	101.15	161.69	95.86	165.6
L13	1200	30	0.04	82.101	163	81.098	163.44	85.14	160.5
L14	1200	30	0.06	148.88	170	159.88	170.19	146.4	173.4
L15	1200	30	0.08	84.231	164	92.524	164.44	81.40	164.8
L16	1200	36	0.04	79.638	157	87.896	157.19	78.50	154.2
L17	1200	36	0.06	123.95	162	125.54	162.96	129.4	167.4
L18	1200	36	0.08	82.217	158	89.28	158.69	81.54	159.2
L19	1400	24	0.04	91.081	166	98.124	166.44	92.74	169.8
L20	1400	24	0.06	145.54	183	149.50	183.44	153.6	179.8
L21	1400	24	0.08	141.42	168	140.32	168.19	143.5	179.1
L22	1400	30	0.04	155.57	181	162.45	181.19	158.3	184.2
L23	1400	30	0.06	172.88	200	180.21	200.44	170.8	197.9
L24	1400	30	0.08	120.5	186	129.98	186.19	124.1	194.7
L25	1400	36	0.04	96.472	170	103.845	170.69	95.72	173.2
L26	1400	36	0.06	152.94	178	160.55	178.69	154.3	179.5
L27	1400	36	0.08	105.9	177	107.8	177.44	107.3	173.7

Therefore, it is evident that both full factorial design and ANN give near accurate solutions. The ANN predicted results were collated with full factorial design of experiment values, leading to high prediction, 95% accuracy. The ANN incorporates the input process parameters that produce the interlock FSWed samples with desirable properties.

Microstructure analysis

Figure 10 shows scanning electron micrograph of AA7075 T7-AA7475 T6 interlock welded joints using FSW technique. Optimum FSW process parameters (L23 conditions) were considered. The microstructure of dissimilar interlock lap joined aluminium alloy is divided into four different phases: the eutectic phase, α-Al, dark coarse primary and intermetallic phases with script shaped structures.²¹ It was clearly evident that primary Al particles which are responsible for improvement in statistical characteristics are heterogeneously distributed in the aluminium matrix.

Figure 10.

Micrographs of different zones on transverse section of dissimilar AA 7075-T6 and AA7475.

Figure 10(a) depicts the breakage of aluminium particles may be attributed to its spine shape with the high aspect ratio and abrading action of the rotating tool. Due to spine like shape, the probability of stress concentration was high due to which led to the formation of fragmented particles.²² It is known that silicon particle present in the dissimilar interlock lap weld is brittle in nature and hence cannot deform plastically like a metallic material. The scanning electron microscopy (SEM) analysis of the parent material confirms the presence of elongated grains and fine intermetallic particulates in both the Al sheets utilized for carrying out FSW. Figure 10(b) shows nugget zone with equiaxed grains in which the size of the grain has been decreased from 5 μm from grain size of 20 μm of weld. Figure 10(c) represents the cross-section microstructure of weld along the weld nugget/thermo mechanical affected zone of welds along the advancing and retreating sides.²³ From the microstructure analysis it is evident that small micro pores and pit holes occurred on the surface of the welds were made using FSW. Figure 10(d) illustrates the high speed FSW tool accelerates the separate grain formation with different grain structures, where TMAZ/NZ are separated by the defined grain boundary in the interlock lap weld.²⁴ Hence due to this fact, the size of grains in thermo mechanically affected zone is much larger than the nugget zone and move upwards between the lap sandwich structure and thereby creating a strong zone between the lap welds. Figure 10(e) represents pit holes and surface deformations due to increased plastic deformation of the rotation tool and excess impact of the tool shoulder on the surface of the weld. It is clear from the microstructure the weld region consists of coarse grains due to low speed of rotating FSW tool and thereby lack of required temperature for obtaining finer grains in the weld region creates the increased plastic deformation which enhances the pile up grains formations in elongated structure in the NZ of the interlock weld.²⁵ The Al-Mg-Si system in AA7075 T6 sheet of interlock weld is the preliminary strengthening phase where β-shaped CFC structure contributes maximum to the strength of the welds. In AA 7475 T7 sheet, presence of the Al-Mg-Cu system performs the strengthening action between the welds. The effect of evolution/fragmentation and distribution of Al particles is represented in Figure 11. It was observed that particles present such as copper, zinc and traces of silicon in aluminium alloys initiated the complete breakage of coarse acicular silicon particles along with significant reduction in aspect ratio.²⁶ It can be observed that the broken components of aluminium alloy particles have dispersed throughout the homogeneous region of nugget zone. It has also been observed in shoulder affected region, the uniform distribution of homogenous particles throughout the nugget zone.²⁷ In SAZ and NZ zones, the agglomeration particles have not accumulated together rather than separated with a fair distance apart.

Figure 11.

Optical micrographs of (a, b) interface at advancing and retreating side respectively, (c) Nugget Zone, and, (d) shoulder affected zone.

Fracture modes and fracture SEM analysis in interlock joints

Figure 12 depicts different fracture mechanisms occurred in FSWed interlock lap joint. Shear and tensile fracture occurred in the interlock FSWed joints, when they are exposed to external load conditions. Welds produced at TRS of 1000 rpm surface crack penetrates at the TMAZ/SZ interface which clearly shows that welds does not have relatable strength to sustain the external force applied. At TRS of 1200 rpm crack creeps in-between the lap weld structures leading to the formation of shear fracture. The shear fracture occurred in between the welds reduced the strength of joints to a greater extent, since the facture arises between the TMAZ/SZ and HAZ.²⁸

Figure 12.

Fracture obtained in Interlock FSW lap weld.

At TRS of 1400 rpm tensile fracture mechanism occurred at the HAZ interface of both the sheets welded. Since HAZ is stronger zone when compared to the TMAZ, the welds made using 1400 rpm have enough strength than the welds made at TRS of 1000 and 1200 rpm, respectively.

Figure 13(a)–(c) represents the fractured AA7075 T6-AA7475 T7 interlock lap welded samples with good, better and best samples made using FSW. The surface of the fractured samples was analyzed using scanning electron microscope for identifying the nature and effect of fracture microstructure on the surface of interlock FSWed samples. A cone shape fracture was formed long the axis of the weld periphery at an angle of 45° during the tensile testing of the weld samples, which led to the formation of micro void coalescence at the fractured surface of the sample.²⁹ The fracture obtained in all the interlock lap weld made at TRS of 1400 rpm (L23) exhibited small and equally spaced dimples which clearly indicates that the fracture mode occurred is ductile in nature. The fractured (L1 and L26) weld samples obtained at tool speeds of 1000 and 1200 rpm exhibited micro porous fractures and AL agglomerations were obtained on the surface of the weld samples. The fracture microstructure of the AA7075 T6-AA7475 T7 made at TRS of 1400 rpm, WS of 30 mm/s and plunge speed rate of 0.06 mm/s (L23) were very fine and uniform when compared with the other welds made using FSW and it was also clearly seen that the formation of micro dimples was very fine and smaller in the sheet kept at the bottom of the lap weld.³⁰ Hence due to this fact, the strength of the weld sample made at 1400 rpm was high than the other welds made using FSW. The interlock weld samples made at TRSs of 1000 and 1200 rpm exposed honey comb structures on the surface of fractured samples with characteristics of plastic deformation. All the fracture of welds made at tool speed of 1400 rpm were homogenous with decreased grain size due to high speed rotation to FSW tool, leading to the effective nixing of materials. The fracture obtained in the best interlock weld sample (L23) is along the heat affected area leading to the softening of material which is made clear by hardness variations and moreover the fracture zone was well within the HAZ of the weld samples.³¹ The location of fracture in friction stir welds plays a crucial role in enhancing the statistical properties of the AA7075 T6-AA7475 T7 interlock lap welds.

Figure 13.

Fractography of interlock FSWed lap weld joints: (a) L1, (b) L26, (c) L23.

Conclusion

The optimum FSW process parameters for joining dissimilar interlock lap weld AA7075 T6-AA 7475 T7 identified by the full factorial design of experiments has been compared with neural network (ANN) to analyse the degree of accuracy of the optimized results. Twenty-seven experiments were carried out using L27 orthogonal array to determine the strength and hardness of the welded samples. Three parameters and three levels considered are TRS, WS and PSR. The analysis of results made by full factorial design of experiments concludes that:

The optimum process parameter for AA7075 T6-AA7475 T7 interlock lap weld joints made by FSW is TRS 1400 rpm, WS 30 mm/s and PSR 0.06 mm/s, respectively.

Full factorial analysis suggests that TRS3, WS2 and PSR2 welding conditions represented the best optimized process parameter conditions which exhibited a maximum tensile strength of (172.88 MPa) and hardness of (200 HV). Variance analysis exhibited TRS as the primary contributor for the strength of the welds by achieving higher tensile and hardness values.

The interlock friction stir lap welding technique accomplished even and uniform distribution of weldment and thereby interlock between the lap joints increased the strength to a considerable amount than the regular lap welding technique.

The TRSs below 1400 rpm and WSs below and above 30 mm/s during the FSW of interlock weld samples exhibited surface cracks, micro voids of increased sizes dimples and AL particle agglomeration.

ANN and full factorial regression models analysed and predicted the tensile and hardness with 95% accuracy levels. The tensile strength and hardness of the experiments made using full factorial regression were compared with ANN predictions.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Rajendran Anand

References

Verma

Misra

. Experimental investigation on friction stir welding of dissimilar aluminium alloys. Proc IMechE, Part E: J Process Mechanical Engineering 2021; 235: 1545–1554.

Khan

Ubaid

Siddiquee

, et al. Microstructural features of friction stir welded dissimilar Aluminium alloys AA2219-AA7475. Mater Res Express 5. Epub ahead of print 2018. DOI: 10.1088/2053-1591/aac4e1.

Anand

Sridhar

. Studies on process parameters and tool geometry selecting aspects of friction stir welding – A review. Mater Today Proc. Epub ahead of print 2019. DOI: 10.1016/j.matpr.2019.12.042.

Tang

Reynolds

. Effect of tool pin features on process response variables during friction stir welding of dissimilar aluminum alloys. Sci Technol Weld Join 2015; 20: 425–432.

Bhushan

Sharma

. Optimization of parameters for maximum tensile strength of friction stir welded AA6082/Si3N4 and AA6082/SiC composite joints. Silicon. Epub ahead of print 2019. DOI: 10.1007/s12633-019-00216-3.

Kadlec

Růžek

Nováková

. Mechanical behaviour of AA 7475 friction stir welds with the kissing bond defect. Int J Fatigue 2015; 74: 7–19.

Pouranvari

. Fracture toughness of martensitic stainless steel resistance spot welds. Mater Sci Eng A 2016; 680: 0–1.

Gachi

Aissani

Baudin

, et al. The microstructure, texture and mechanical properties of friction stir welded aluminum alloy. Russ J Non-Ferrous Met 2020; 61: 523–533.

Anand

Sridhar

. Microstructure and mechanical properties of interlock friction stir weld lap joint AA7475-T7 using fractional factorial design. Proc IMechE, Part C: J Mechanical Engineering Science 2022; 236: 318–329.

10.

Kwee

. Weldability of high-strength aluminium alloy EN AW-7475-T761 sheets for aerospace applications, using refill friction stir spot welding. Weld World 2019; 63: 1001–1011.

11.

Padmanabhan

RAR

. Interlock friction stir weld of dissimilar aluminum alloys AA 7075 T6-AA7475 T7 using a novel asymmetric dual-pin tool: mechanical properties and microstructure characterization. Metallogr Microstruct Anal 2022; 11: 225–233.

12.

Natrayan

Senthil Kumar

. An integrated artificial neural network and Taguchi approach to optimize the squeeze cast process parameters of AA6061/Al2O3/SiC/Gr hybrid composites prepared by novel encapsulation feeding technique. Mater Today Commun 2020; 25: 101586.

13.

Ahmed

Nagesh

Raju

, et al. Studies on the effect of welding parameters for friction stir welded AA6082 reinforced with Aluminium Oxide. Mater Today Proc 2020; 20: 108–119.

14.

Babu

Elangovan

Balasubramanian

, et al. Optimizing friction stir welding parameters to maximize tensile strength of AA2219 aluminum alloy joints. Met Mater Int 2009; 15: 321–330.

15.

Salih

Neate

, et al. Influence of process parameters on the microstructural evolution and mechanical characterisations of friction stir welded Al-Mg-Si alloy. J Mater Process Tech 2020; 275: 116366.

16.

Natrayan

Ravichandran

Veeman

, et al. Enhancement of mechanical properties on novel friction stir welded Al-Mg-Zn alloy joints reinforced with Nano-SiC particles. J Nanomater 2021. Epub ahead of print 2021. DOI: 10.1155/2021/2555525.

17.

Bhuiya

MMK

Rasul

Khan

MMK

, et al. Optimisation of oil extraction process from Australian native beauty leaf seed (Calophyllum Inophyllum). Energy Procedia 2015; 75: 56–61.

18.

Khalkhali

Sarmadi

Sarikhani

. Investigation on the best process criteria for lap joint friction stir welding of AA1100 aluminum alloy via Taguchi technique and ANOVA. Proc IMechE, Part E: J Process Mechanical Engineering 2017; 231: 329–342.

19.

Soundararajan

Ramesh

Sivasankaran

, et al. Modeling and analysis of mechanical properties of aluminium alloy (A413) processed through squeeze casting route using artificial neural network model and statistical technique. Adv Mater Sci Eng 2015. Epub ahead of print 2015. DOI: 10.1155/2015/714762.

20.

Senapati

Panda

Bhoi

. Prediction of multiple characteristics of Friction-Stir welded joints by Levenberg Marquardt algorithm based artificial neural network. Mater Today Proc 2019; 41: 391–396.

21.

Qin

Zhao

Zhang

, et al. Microstructures and mechanical properties of Al-Mg2Si-Si alloys resistance spot welded with Al-Si interlayers. J Mater Res Technol 2019; 8: 4318–4332.

22.

Rajendran

Srinivasan

Balasubramanian

, et al. Effect of tool tilt angle on strength and microstructural characteristics of friction stir welded lap joints of AA2014-T6 aluminum alloy. Trans Nonferrous Met Soc China 2019; 29: 1824–1835.

23.

Çevik

Ozçatalbaş

Gülenç

. Rührreibschweißen einer 7075-T651-Aluminiumlegierung. Prakt Metallogr Metallogr 2016; 53: 6–23.

24.

Jacob

Maheshwari

Siddiquee

, et al. Improvements in strength and microstructural behaviour of friction stir welded 7475 aluminium alloy using in-process cooling. Mater Res Express 2018; 5: 076518.

25.

Liu

Peng

, et al. The effect of interface defect on mechanical properties and its formation mechanism in friction stir lap welded joints of aluminum alloys. J Mater Process Technol 2016; 238: 244–254.

26.

Jafari

Mansouri

Honarpisheh

. Investigation of residual stress distribution of dissimilar Al-7075-T6 and Al-6061-T6 in the friction stir welding process strengthened with SiO2 nanoparticles. J Manuf Process 2019; 43: 145–153.

27.

Rezaei

Mirbeik

Bisadi

. Effect of rotational speeds on microstructure and mechanical properties of friction stir-welded 7075-T6 aluminium alloy. Proc IMechE, Part C: J Mechanical Engineering Science 2011; 225: 1761–1773.

28.

Paidar

Ojo

Moghanian

, et al. Modified friction stir clinching with protuberance-keyhole levelling: a process for production of welds with high strength. J Manuf Process 2019; 41: 177–187.

29.

Goel

Khan

, et al. Investigation on material mixing during FSW of AA7475 to AISI304. Mater Manuf Process 2019; 34: 192–200.

30.

Gao

, et al. Effect of pin length and welding speed on lap joint quality of friction stir welded dissimilar aluminum alloys. Int J Adv Manuf Technol 2018; 98: 1461–1469.

31.

Hajihashemi

Shamanian

Niroumand

, et al. Microstructure and mechanical properties of Al-6061-T6 alloy welded by a new hybrid FSW / SSW joining process microstructure and mechanical properties of Al-6061-T6 alloy welded by a new hybrid FSW / SSW joining process. Sci Technol Weld Join 2016; 1718: 493–503.

An integrated ANN and design of experiments technique to optimize the FSW input parameters of novel interlock lap weld

Abstract

Keywords

Introduction

Materials selection and preparation

Full factorial design of experiments

Statistical characterizations

Results

Statistical analysis

Analysis of variance

Regression equation

Artificial neural network

Microstructure analysis

Fracture modes and fracture SEM analysis in interlock joints

Conclusion

Footnotes

Declaration of conflicting interests

Funding

ORCID iD

References