P RODUCTION E NGINEERING A RCHIVES

The paper investigates the modelling and optimization of the notch-repaired/friction stir stitched AISI 201 stainless steel welds via the use of a non-consumable tool-based repair process. The repair process employs a sequential hopping-stitching approach. This approach involves the application of two inter-cepted and completely overlapped plunging actions of a probe-less titanium carbide tool to create an effective refilling and repair of the notched zone. Box-Behnken design (BBD) was employed for the experimental planning, modelling, and optimization of the notch-repair process. Tool rotational speed, penetration depth and dwell time of the tool were the studied process parameters while tensile strength was the response variable. A quadratic model was identified as the best model for the notch-repaired welds based on the combination of a low sequential P-value of 0.008216, a high lack of fit P-value of 0.931366, and a close to unity adjusted and predicted R-square values. The process parameter and their interaction effects on the tensile strength of the repaired notch were identified via the ANOVA analysis. Plunge depth (main effect) and interaction effect of tool rotational speed and dwell time had significant influences on the notch-repair process and the resultant tensile strength of the AISI 201 stainless steel. The visual representations of these effects were shown through the 2D elliptical contour and 3D response surface plots. The optimized process parameters were identified as 1215.9795 rpm, 0.40262212 mm, and 5.98706376 s while the resultant notch-repaired joint yielded a tensile strength of 886 MPa, which is close to the predicted value.


Introduction
Repair of damaged parts in offshore (subsea/underwater) and onshore environments is requisite to prolong the service/useful life of structural parts, reduce part-replacement cost, and fabrication lead-time of new parts (Zhang et al., 2016). The conventional underwater/wet arc welding methods are known to be prone to hydrostatic pressure-induced arc instability once the depth at which welding/repair is mandatory goes beyond 40 m (Zhang et al., 2016). Also, an additional operational cost is usually incurred in welding/repairing parts at underwater depths beyond 40 m as a dry chamber/cabin is required to ensure the safety of welders/divers (Zhang et al., 2016). Thus, there is a need to have a repair strategy or method that will be suitable for underwater repairs irrespective of the level of underwater depths. Solid-state repair approaches (such as friction stir stitching and friction taper plug welding) are being considered as viable alternatives to the arc welding/repair process in the modern industrial drive.
The concept of keyhole repairing in friction stir spot welding process can be adapted for structural repairs. Reimann et al. (2017) investigated the keyhole repairing of the AA7075-T651 welds via the use of a refill friction stir spot welding process. Aliakbari et al. (2018) employed the friction stir processing technique to repair fusion welds of AA6061 alloy. It was reported that weld defects were majorly eliminated through the application of the friction stir process. Li et al. (2011) employed the overlapping-friction stir welding approach to repair AA2219-T6 Al alloy. It was revealed that abnormal coarsening of Al2Cu particles ensued in the sample due to phenomena like aggregation mechanism, diffusion mechanism "I" and diffusion mechanism "II".
Recently, researchers have investigated friction forging (Derazkola et al., 2020), additive friction stir deposition (Perry et al., 2020), and friction stir-based underwater welding processes (Rouzbehani et al., 2018) (Guiju Zhang et al., 2021. The success of these studies in an underwater environment has spurred the adaptation of solid-state friction stir processing for repair purposes. Friction hydro-pillar processing, underwater friction taper plug welding, or friction stitch welding process has been employed for repair purposes in literature (Zhang et al., 2016) (Teng et al., 2017). This approach involves a combination of drilling (of a blind hole) and filling (of the drilled hole) by employing a consumable tapered plug mounted/attached to the end of the rotating welding tool. The plugs are overlapped in a sequential run to establish a good repair. The downward force, induced plastic deformation, and rotating effects of the consumable plug facilitate the filling of the drilled hole. Wang et al. (2019) reported that the continuously extruded viscoplastic material filled up the pre-drilled hole along the arc welded DH36 joint. This solid-state repair approach is independent of underwater pressure and depths. As a result, it is considered a viable approach for repairs in an underwater environment. Teng et al. (2017) carried out the friction stitch welding/repairing of arc welded DH36 joints in an underwater condition and a high-quality repair was obtained.
Summarily, the use of a consumable plug-assisted tool has been utilized for repair purposes in literature. A somewhat amount of time, cost, and drudgery go into the plug (tool) preparation, and the hardness and weldability of the plug material also play a significant role in achieving an effective repair. Zhang et al. (2016) acclaimed that the use of a plug with lower hardenability and better weldability is essential in achieving good underwater friction stitch welds. In view of this, this paper examines the use of a non-consumable tool to repair gas metal arc welded line of AISI 201 austenitic stainless steel via friction stir stitching approach. This steel is useful in nuclear, houseware, and chemical processing applications (Filho et al., 2016(Filho et al., , 2017(Filho et al., , 2019. Modeling and optimization of the friction stir stitched welds (notch-repaired welds) were carried out by using the Box-Behnken design.

Materials and method
The materials used for this study are a 5 mm thick AISI 201 stainless steel (SS) plate, 2 mm diameter 308LSi filler wire, and an 8 mm diameter titanium carbide tool. The elemental compositions and the mechanical properties of these materials are shown in Tables 1 and 2 respectively. The as-received AISI 201 SS plates were cleaned and cut into dimensions of 1550 mm × 70 mm for butt welding. The abutting edges/ends of two plates were prepared/chamfered to allow for the deposition of molten filler metal during the welding process. Gas metal arc welding (GMAW) of the prepared AISI 201 SS plates with a 308LSi filler wire was carried out using a 32 V welding voltage, 180 A welding current, and 350 mm/min welding speed based on the optimized works of Kapil et al. (2014). Carbon dioxide shielding gas was utilized for the welding process due to its cheap cost and similar penetration effect as compared to helium gas. The as-welded plates (1550 mm × 140 mm) were cut across the welded butt-path into rectangular samples (30 mm × 140 mm) with the aid of a guillotine machine (see the cut samples in Fig.1) The cut samples (shown in Fig.1) were subjected to further preparation before the friction stir stitching (FSS) process. A 2 mm deep blank hole or notch was created at the center of the cut samples along the welded butt line with the aid of a 5 mm drill bit.

Fig. 1. Cut samples of the as-welded AISI 201 SS plates
The notched sample was rigidly clamped to the worktable of the milling machine (adopted for the stitching process) with the aid of fixtures before the repair process. The established notch was repaired with the aid of a non-consumable probeless titanium carbide tool having 8 mm diameter via a 3-sequential step approach (see Fig. 2) under the adequate supply of freshwater to the tool-material contact (red-hot) zone. The repair process involves two intercepted plunging and a completely overlapped plunging action of the welding tool to create a notch-repaired zone/friction stir stitched weld. This hopping sequence of stitching was adopted to ensure effective refilling and repair of the notched zone.
The plasticized material underneath the tool aided the refilling of the notched zone during the two intercepted plunging actions) while the consolidated and repaired zone was established by the third overlapped plunging action of the tool. Fig.3 shows the surface appearance of a repaired sample.

Fig. 3. A sample showing the repaired notch region
The experimental notch-repair process was planned for by using Box-Behnken design (BBD), a type of response surface methodology (RSM), to reduce the number of experimental runs and costs. The BBD is a 3-level design and the selected process parameters for this study include tool rotational speed, plunge depth and dwell time. The actual and coded levels of the selected process parameters are provided in Table 3. A 15experimental run BBD was employed for this study, while the tensile strength of the repaired samples was employed as the quality characteristic for investigating the designed experiment. The tensile samples were fabricated from the repaired samples (shown in Fig.3) and the resultant tensile strength of the samples was obtained in accordance with ASTM E8 standard using a SANTAM 150 tensile testing machine at a cross-head speed of 5 mm/min. The average of three tensile strengths is reported as the actual tensile strength in this paper. However, the first step in the RSM analysis is to identify a suitable function/approximation of the real relation between the tensile strength (y) and the input parameters (x) (Vahdati and Moradi, 2020). These functions are expressed in the form of linear and quadratic models given as Eq. 1 and 2. Where 0 , , , , k and are the constant value, the first order linear coefficient, the second-order quadratic coefficient, the interaction coefficient, the number of independent process parameters, and the error respectively. The analysis of variance (ANOVA) was employed to check the main interaction and quadratic effects in the Box-Behnken design.

Results and discussion
Solidification-induced heterogenous structure typical of fusion welds constitutes the weld metal of the GMAW'ed AISI 201 stainless steel (see Fig.4a) while fine structure is found in the friction stir stitched part of the joint (see Fig.4b). The observed structure in Fig.4b is due to the severe plastic deformation and dynamic recrystallization effect induced by the rotating titanium carbide tool on the AISI 201 substrate.
The tensile strengths of the repaired samples are shown in Table 4. Design-Expert software was employed to perform model fitting based on the Box-Behnken experimental design in Table 4. According to the model/fit summary shown in Table 5, a quadratic model, given as Eq. (1), is suggested as the best model for this experimental study due to a combination of a low sequential P-value of 0.008216 (< 0.05), a high lack of fit P-value of 0.931366 (not significant relative to error), and a close to unity adjusted and predicted R-square values. The mathematical expression provided in Eq. (3)  The ANOVA result further validates that the quadratic model chosen from the model/fit summary is significant in Table 6. The high Fisher's F-value with a low probability value (< 0.05 p-value) of the quadratic model indicates that the model is significant while the non-significant lack of fit Fvalue implies that good predictability can be obtained with the model. Ghelich et al. (2019) reported that the quality of the quadratic (polynomial) model and its general predictive/prognostic capability can be ascertained by the coefficient of determination (R2 value). Azizi et al. (2012) reported that a model with a good fit should have at least an R2 value of 0.80. The R2, from the ANOVA results, was found to be 0.958 for the friction stir stitched joints. This observation indicates that an acceptable relationship exists between the actual and the predicted values of the joint strength. Similarly, the ANOVA results reveal whether the variables/terms associated with the model are significant based on the variability of the obtained experimental results (in Table 6).   Table 6, the significant model terms are one independent variable B, one interactive cross-product coefficient of AC, and two quadratic term coefficients of A², and B² (having < 0.05 p-value). These significantly affect the average tensile strength of the friction stir stitched joints whereas the other terms (that are removed from the model description) have an insignificant effect on the tensile strength of the joint. The difference between the predicted R² (0.8098) and adjusted R² (0.8831) is less than 0.2. This shows that a reasonable agreement exists between them for good model predictability. Also, one of the notable diagnostic plots is checked to view any deficiency in the model fitting (Soleymani et al. 2015). The graphical representation of the predicted vs. actual values is shown in Fig.5. The closeness of the points to the diagonal line (in Fig.5) further indicates good predictability of the developed quadratic model in predicting the tensile strength of the friction stir stitched AISI 201 joints. The relationship between the independent process variables/parameters and the response variable (tensile strength) is visually appraised via the 2D elliptical contour (see Fig.6a) and 3D response surface plots (see Fig.6b). According to the ANOVA result, one interactive cross-product coefficient of AC is significant in Table 5. Thus, the mutual interaction/pairwise combination of two variables (tool rotational speed and dwell time) while the other variable (plunge depth) is held at its center point (zero levels) is revealed in Fig.6. Two regions show improvement in the tensile strength of the friction stir stitched joints in the elliptical contours (see Fig.6a) due to the interactive effect. The first region is the interactive effect of a low level of tool rotational speed and the highest level of dwell time (see the yellowish-orange coloration towards the dwell time in Fig.6a). Although the highest tensile strength (red contour) is not attained at a low tool rotational speed, the thermally-induced frictional heat owing to the prolonged exposure time is somewhat significant in this case.
The second region shows the pairwise combination of high tool rotational speed and low dwell time (see the red contour in Fig.6a). The increase in the tool rotational speed at a reduced/low level of dwell time is desirable in improving the tensile strength of the friction stir stitched joints in Fig.6a and b. A direct relationship exists between the tool rotational speed and inherent heat input during friction stir-based processing (Ojo, 2019) (Heydari et al. 2019). This indicates that the induced heat input (deformation and friction) is significant in improving the load-bearing attribute of the repaired or friction stir stitched joints at a high level of tool rotational speed.
The optimization of the friction stir stitching process was carried out based upon the desirability function. The target objective of the process was to maximize the tensile strength of the repaired/stitched joint by recalculating all accountable factors using the desirability functions. Table 7 shows the criteria/constraints employed for optimizing the process. All the independent variables were in ranges based on the designed experiment. The lower and upper limits of the factors (A, B, and C) were obtained from the Box-Behnken experimental design levels. According to the goal and settings (in Table 7), the predicted tensile strength of 915.15 MPa at optimized values of 1215.9795 rpm, 0.40262212 mm, and 5.98706376 s is obtained.  7 shows the desirability ramp obtained from the numerical optimization of the friction stir stitching process of the AISI 201 SS plates. Fig.8 also reveals the 2D elliptical contour showing the desirability and tensile strength of the friction stir stitched joints. The desirability of 1 (unity) is obtained at a high level of dwell time in the optimized result while the plunge depth and tool rotational speeds are kept at somewhat low levels. Fig.8 further validates that desirability is more suitable in obtaining the optimum parameter levels as compared to the contour plot of the tensile strength. The accuracy of the optimization was checked by performing a confirmatory experiment. The control setting on the milling machine was not designed to set a rotational speed level of 1216 rpm. Consequently, the combination of tool rotational speed of 1200 rpm, plunge depth of 0.403, and dwell time of 5.99 s was employed as the optimized setting for the validation check. The adjusted/optimized experimental setting produced a tensile strength of 886 MPa while the predicted value was 915.15 MPa. The difference in the values is attributed to the adjustment in the tool rotational speed during the validation experimental run.
The fracture location of the friction stir stitched AISI 201 stainless steel welds occurs through the weld centres as shown in Fig. 9a. This is owing to the heterogeneous microstructure induced by the gas metal arc welding process on the AISI 201 stainless steel. This structure induces stress concentration at the weld metal of the weld samples during the tensile testing process. The continuous axial loading of the friction stir stitched sample initiates crack at the fusion part of the joint and eventually propagates into the friction stir stitched part of the joint. However, the assessment of the fracture surfaces (in Fig.9b-e) reveals different degrees of dimples. This observation indicates that ductile fracture mode is the dominant fracture mode in the friction stir stitched AISI 201 stainless steel joints.

Conclusion
This study successfully applied a non-consumable tool to repair the gas metal arc welded line of AISI 201 stainless steel via the use of a friction stir stitching method to eliminate drudgery and cost associated with tool preparation. The modeling and optimization of the friction stir stitched AISI 201 stainless steel welds were carried out. The findings of this study are summarized as follows: i. A non-consumable tool-based friction stir stitching is effective for the weld line repairs of AISI 201 stainless steel, provided that sequential hopping-stitching manna is adopted.
ii. Box-Behnken design methodology and quadratic expression are effective for optimizing and modelling the friction stir stitching process respectively. iii. The optimized process parameters for the stitching process are tool rotational speed, plunge depth and dwell time of 1215.9795 rpm, 0.40262212 mm, and 5.98706376 s respectively. iv. The optimum tensile strength of the friction stir stitched sample is 886 MPa v. Plunge depth had the dominant impact on the friction stir stitching process while an interactive effect existed between the rotational speed and dwell time of the tool.